Prosecution Insights
Last updated: November 16, 2026
Application No. 17/436,623

SYSTEM FOR ACOUSTIC IDENTIFICATION OF OBSTRUCTION TYPES IN SLEEP APNOEA, AND CORRESPONDING METHOD

Non-Final OA §103
Filed
Sep 06, 2021
Priority
Mar 07, 2019 — DE 10 2019 105 762.0 +1 more
Examiner
RAMESH, TIRUMALE K
Art Unit
2121
Tech Center
2100 — Computer Architecture & Software
Assignee
Diametos GmbH
OA Round
4 (Non-Final)
28%
Grant Probability
At Risk
4-5
OA Rounds
52%
With Interview

Examiner Intelligence

Grants only 28% of cases
28%
Career Allowance Rate
14 granted / 51 resolved
-27.5% vs TC avg
Strong +25% interview lift
Without
With
+24.6%
Interview Lift
resolved cases with interview
Typical timeline
Avg Prosecution
23 currently pending
Career history
91
Total Applications
across all art units

Office Action

§103
DETAILED ACTION Notice of Pre-AIA or AIA Status The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . Response to Arguments (Submitted on 6/24/2026) Regarding 103 rejections The applicant made no amendments. The examiner ascertains that the present focus of the invention is to use three classifiers in which the first classifier is the classification of snoring noise origin, the second classifier is the classification of the respective mouth position, and the third classifier is based on fusing (combining) the first classification and second classifier and that the third classifier outputs the respective obstruction type. The applicant argues on Page 9 that the reference “Janott” says nothing about obstruction types. Further, the applicant argues on Page 10 that the reference “Janott” is silent on the third classification that provides classification of an obstruction type derived from snoring noise origin and a mouth position signal. Examiner’s Response The examiner disagrees with the applicant about the argument that the reference “Janott” says nothing about obstruction types. First, the examiner identifies that the obstruction types can caused by of “nasal snoring” caused by narrowing in or around the nasal passages, which is taught by “Janott” in [1.4, Page 107], “mouth snoring” caused by narrowing in the mouth or oral cavity which is taught by “Janott” in [6.2, Page ], “soft palate snoring” caused by narrowing the software palate which is taught by “Janott” in [1.1, Page 106]. The examiner after further review of the application with consultation with a primary examiner in the “medical” area art unit with the feedback that the invention may be well known problem area in the medical area. Based on the consultant feedback, the examiner submits that the reference “Alshaer” used in the dependent claims does indeed teaches the “third classifier” in [0081] and the reference also provided substantial support in other paragraphs of the prior art. As the applicant has made no amendments, the examiner has used new grounds of rejection (not necessitated by the amendment) that is strongly supported by NPL references published several years ago, and as a result, the examiner determined that a new NON-FINAL REJECTION was necessitated for the application. The examiner uses two new NPL references “Nwe” and “Senny”. Reference “Nwe” teaches the first reference and fused classification (third classification ) and reference “Senny” teaches the mouth position classification (Second classification). In CONCLUSION, the examiner rejects claims 1-12, 14-22, 24-25, and 27-34 under 103 as NON- FINAL REJECTION. Claim Rejections - 35 USC § 103 The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. Claims 1, 3-4, 6, 8-12, 21, 28-31, and 34 are rejected under 35 U.S.C. 103 as being unpatentable over Tin Lay Nwe et.al (hereinafter Nwe) An Integrated Solution for Snoring Sound Classification Using Bhattacharyya Distance based GMM Supervectors with SVM, Feature Selection with Random Forest and Spectrogram with CNN, INTERSPEECH 2017, August 20–24, 2017, In view of Avi Kopelman et.al. (hereinafter Kopelman) US 2016/0199215 A1, in view of Frédéric Senny et.al (hereinafter Senny) Midsagittal Jaw Movement Analysis for the Scoring of Sleep Apneas and Hypopneas, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 1, JANUARY 2008. Regarding claim 1: (Currently Amended) Nwe discloses: - a) receiving, through an input interface, a snoring-noise signal comprising noise of obstructive sleep apnea [Abstract, 3467]: Snoring is caused by the narrowing of the upper airway and it is excited by different locations within the upper airways. This irregularity could lead to the presence of Obstructive Sleep Apnea Syndrome (OSAS). Diagnosis of OSAS could therefore be made by snoring sound analysis. [4.2, Page 3469]: we employ CNN for snore sound classification. Sounds have global spectral pat terns, as well as local properties such as being more transient or smoother in the time-frequency domain [14]. These can be exposed by using a model that exploits locality leading us to explore two different feature extraction strategies in the context of Convolutional neural networks (CNN) based deep learning. Hence, we employ CNN, a state of the art 2D feature extraction model, to exploit local structures, with log power spectrogram as input in our 3rd sub-system for snore sound classification. - b) generating, using a first classifier, a snoring-noise origin classification, wherein the first classifier is adapted to learn in a first training mode, when a first plurality of snoring-noise signals is input with a corresponding type of snoring-noise origin, such that in a first identification mode, [Abstract, Page 3467]: Snoring is caused by the narrowing of the upper airway and it is excited by different locations within the upper airways. This irregularity could lead to the presence of Obstructive Sleep Apnea Syndrome (OSAS). Diagnosis of OSAS could therefore be made by snoring sound analysis. [4.2, Page 3469]: we employ CNN for snore sound classification. Sounds have global spectral pat terns, as well as local properties such as being more transient or smoother in the time-frequency domain [14]. These can be exposed by using a model that exploits locality leading us to explore two different feature extraction strategies in the context of Convolutional neural networks (CNN) based deep learning. Hence, we employ CNN, a state of the art 2D feature extraction model, to exploit local structures, with log power spectrogram as input in our 3rd sub-system for snore sound classification. [BRI: the 3 rd sub-system is the first classifier] - the first classifier is configured to generate the snoring-noise origin classification for a particular snoring-noise signal from a group of predefined types of snoring- noise origin, wherein each of the predefined types of snoring-noise origins specifies a predefined location in a subject's head, wherein each of the predefined types of snoring-noise origins specifies a predefined location in a subject's head; [5, Page 3469]: We use MUNICH-PASSAU SNORE SOUND CORPUS (MPSSC) which is provided by ComParE2017 challenge for snore sound classification based on their excitation location. Four classes are defined based on the VOTE scheme. V refers to Velum (palate), including soft palate, uvula, lateral velopharyngeal walls. O is for Oropharyngeal lateral walls which include palatine tonsils. T is for Tongue, including tongue base and air way posterior to the tongue base. And, E is for Epiglottis. [BRI: “VOTE” is a standardized classification system that defines four anatomical regions of the upper airway, used to identify the origin (site) of snoring and obstructive sleep apnea, and is applied in both endoscopic evaluation and snoring sound analysis] [2, Page 3468]: Four types of snore sounds considered correspond four different excitation locations within upper airways. [1, Page 3467]: Snoring occurs when there is a narrowing or obstruction in the upper airways, the turbulence of air and the vibrations of soft palate [1]. Characteristics of the snore sound change depending on the intensity of the vibration on different locations (or excitation location) within the upper airway. [BRI: in the VOTE snore-source classification scheme, each predefined type is explicitly tied to a specific anatomical region of the upper airway. These are not arbitrary acoustic classes but they correspond to distinct, predefined excitation locations in the subject’s head where the vibration or obstruction producing the snore originates. Snoring that results from narrowing or obstruction in the upper airway can indeed originate from a specific, predetermined anatomical site in the head and neck, but the exact location varies between individuals and can change during sleep] - d) generating, using a third classifier, an obstruction type classification from a group of predefined obstruction types of obstructive sleep apnea, wherein the third classifier is adapted to identify in a third identification mode, the obstruction type classification for a particular snoring- noise origin classification identified by the first classifier and a particular mouth position classification identified by the second classifier are input, wherein at least some of the predefined obstruction types specify constriction at one of the predefined locations. [2.2, Page 3468]: As for the speech signal, it is generated by vibrations of the vocal cords, which is filtered and transmitted through the upper airway and oral cavity [5]. [Introduction, Page 3467]: A study which classifies snoring sound based on excitation location is classification between two types of snore sounds: oral and nasal snoring sounds [8]. This study uses fundamental frequency and the maximum of the amplitude spectrum in the specific band to classify two types of snoring sounds. However, there is still a shortage of robust methods to automatically detect snoring sounds based on their excitation location. The treatment will be effective if it is targeting the area in the upper airways where the snoring sound is generated. In this paper, we propose 3 sub-systems for classifying 4 types of snore sounds based on their excitation location. We then fuse these 3 sub-systems to generate final classification results. [1, Page 3467]: We employ machine learning methods such as Support Vector Machine(SVM), Random Forest (RF) and Convolutional Neural Network(CNN) as classifiers for the three sub-systems respectively. System block diagram of our snore sound classification is shown in Figure 1. PNG media_image1.png 268 643 media_image1.png Greyscale [6, Page 3469]: We conduct several experiments to investigate the effectiveness of the individual sub-systems and fused system. Firstly, we examine effect of integrating Bhattacharyya based GMM super vector to the baseline feature set. We use Support Vector Ma chine(SVM) for classification with the integrated feature set. Secondly, we investigate feature selection method, CFS, to improve the performance of the baseline system by using a subset of baseline feature set which is important for classification of 4 different snore sounds, [6.1, Page 3470]: We use training set to train the snore sound models. The result in Table 1 shows that snore sound classification accuracy improves 7.7% absolute over the challenge baseline system for development set (3rd row of Table 1) when we integrate Bhattacharyya based GMM supervectors (Bhat-GMM-Sup) to baseline features, PNG media_image2.png 283 639 media_image2.png Greyscale Nwe do not explicitly disclose: - system for identification of obstruction types in sleep apnea, the system comprising: one or more processors; and memory storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: However, Kopelman discloses: - system for identification of obstruction types in sleep apnea, the system comprising: one or more processors; and memory storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: [0002]: Obstructive sleep apnea (hereinafter “OSA”) is a medical condition characterized by complete or partial blockage of the upper airway during sleep. The obstruction may be related to relaxation of soft tissues and muscles in or around the throat (e.g., the soft palate, back of the tongue, tonsils, uvula, and pharynx) during sleep. OSA episodes may occur multiple times per night and disrupt the patient's sleep cycle. Suffers of chronic OSA may experience sleep deprivation, excessive daytime sleepiness, chronic fatigue, headaches, snoring, and hypoxia. [0039]: Patients suffering from sleep apnea may experience restricted airflow due to blockage of the upper airway if the upper and lower jaws 100, 102 remain in their habitual occlusal relationship during sleep due to relaxation of soft tissues in or around the upper airway. [0006]: In one aspect, a system for monitoring and treating sleep apnea in a patient is provided, the system comprising: one or more sensors configured to monitor the patient for symptoms associated with sleep apnea; an intraoral appliance wearable by the patient; one or more processors; and memory comprising instructions executable by the one or more processors to cause the one or more processors to: receive a set of sensor data from the one or more sensors, detect, using a machine learning algorithm, onset of a sleep apnea event based on the set of sensor data, and transmit a control signal to the intraoral appliance to cause the intraoral appliance to displace a lower jaw of the patient from a first position to a second position in order to treat the sleep apnea event. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA and first classifier. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Nwe and Kopelman do not explicitly disclose: - c) generating, using a second classifier, a mouth position classification, wherein the second classifier is adapted to learn in a second training mode, when a second plurality of snoring-noise signals is input with a corresponding mouth position, such that in a second identification mode, the second classifier is configured to generate the mouth position classification for a particular snoring-noise signal from a group of predefined mouth positions; However, Senny discloses: - c) generating, using a second classifier, a mouth position classification, wherein the second classifier is adapted to learn in a second training mode, when a second plurality of snoring-noise signals is input with a corresponding mouth position, such that in a second identification mode, the second classifier is configured to generate the mouth position classification for a particular snoring-noise signal from a group of predefined mouth positions; [1, Page 87]: A sleep apnea occurs when breathing stops while sleeping. The clinical definition of an apnea is a decrease in magnitude of the respiration movements lower than 20% of the reference value for at least 10 s. An apnea is either obstructive (OA), central (CA), or mixed (MA) depending on whether respiratory effort is completely present, absent, or partially present during the apnea. [Abstract, Page 87]: given the importance of the detection and classification of sleep apneas and hypopneas (SAHs) in the diagnosis and the characterization of the SAH syndrome, there is a need for a reliable noninvasive technique measuring respiratory effort. This paper proposes a new method for the scoring of SAHs based on the recording of the midsagittal jaw motion (MJM, mouth opening) and on a dedicated automatic analysis of this signal [I, Introduction, Page 88]: The topic of this paper is the automatic detection and classification of SAHs through the midsagittal jaw motion (MJM), a novel Pes surrogate. The rationale is based on the principle that when the upper airway collapses and when, in the meantime, the pressure in the lungs becomes more and more subatmospheric, the jaw is passively lowered by thoracic attraction called thoracic tugs [27]. The consequent opening of the mouth reflects, during sleep, a stronger effort to breath. [I, Introduction, Page 88]: Fig. 1 depicts eight patterns of the MJM signal in cases of (a) OA, (b) CA, (c) MA, (d) H, (e) healthy sleep, (f) snoring, (g) awake and (h) isolated artefacts (swallowing, etc.). To sum up the previous findings [27]–[31], the three main features of the MJM signal to suspect and score SAHs are as follows: 1) a significant opening of the mouth, 2) an SJM occurring after the sleep event (pointed out by an arrow in Fig. 1), and 3) the oscillating jaw movements related to respiratory effort which are expected to be present during an OA. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA. Senny teaches the second classifier (mouth position). One of ordinary skills would be motivated to combine Nwe , Kopelman and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Regarding claim 3: (Currently Amended) Nwe and Kopelman do not explicitly disclose: - the first classifier and the second classifier being adapted such that the respective training of the first classifier and the second classifier with an additional plurality of snoring-noise signals takes place together and simultaneously, the respective snoring-noise signal used including the respective type of snoring-noise origin and the respective mouth position as corresponding information. However, Senny discloses - the first classifier and the second classifier being adapted such that the respective training of the first classifier and the second classifier with an additional plurality of snoring-noise signals takes place together and simultaneously, the respective snoring-noise signal used including the respective type of snoring-noise origin and the respective mouth position as corresponding information. [I, Page 88]: Fig. 1 depicts eight patterns of the MJM signal in cases of (a) OA, (b) CA, (c) MA, (d) H, (e) healthy sleep, (f) snoring, (g) awake and (h) isolated artefacts (swallowing, etc.). [IV, Page 94]: Classification of SAHs was per formed using MLPs with local and contextual features. [IV, Page 94]: The classification of SAHs by jaw motion analysis is quite accurate with an agreement of 73.1%. Such results were achieved by the following: 1) decomposing the four-classes problem into easier binary subproblems, 2) ordering their application in descending priority (CA recognition is more important than the separation between H and OA; Fig. 6), and 3) viewing obstructive hypopneic and apneic events as one kind of obstructive respiratory events, [I, Page 88]: The topic of this paper is the automatic detection and classification of SAHs through the midsagittal jaw motion (MJM), a novel Pes surrogate. [I, Page 88]: Fig. 1 depicts eight patterns of the MJM signal in cases of (a) OA, (b) CA, (c) MA, (d) H, (e) healthy sleep, (f) snoring, (g) awake and (h) isolated artefacts (swallowing, etc.). [I, Page 88]: the three main features of the MJM signal to suspect and score SAHs are as follows: 1) a significant opening of the mouth, 2) an SJM occurring after the sleep event (pointed out by an arrow in Fig. 1), and 3) the oscillating jaw movements related to respiratory effort which are expected to be present during an OA. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA. One of ordinary skills would be motivated to combine Nwe, Kopelman and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Regarding claim 4: (Previously Presented) Nwe discloses: - the first classifier being adapted to identify, indicate and forward to the third classifier in the identification mode, the respective type of snoring-noise origin with respective probability. [2.2, Page 3468]: GMM-supervector with Bhattacharyya based kernel Bhattacharyya distance is a separability measure between two Gaussian distributions as in equation (2). In equation (2) Σa i and Σb i are the covariance matrices of the sounds adapted from Universal Background Model (UBM). And, mi a and mi b are the adapted mean vectors. Σu i is the covariance matrix of the UBM. pa and pb are the probabilistic models, GMMa and GMMb, respectively. The first term of equation (2) gives the class separability due to the difference between class means, while the second term gives the class separability due to the variance between class covariance. Based on the first two terms, Bhattacharyya distance based kernel and Bhattacharyya distance based GMM supervector are formulated in [9]. The ith sub vector of Bhattacharyya distance based GMM-supervector for a snore sound is shown in equation (3) [9]. Regarding claim 6: (Previously Presented) Nwe discloses: - the mouth position; identified by the second classifier and an obstruction type are input, such that in the identification mode, it identifies the input obstruction type as the most probable obstruction type with the respective type of snoring-noise origin and the respective mouth position [5, Page 3469]: We use MUNICH-PASSAU SNORE SOUND CORPUS (MPSSC) which is provided by ComParE2017 challenge for snore sound classification based on their excitation location. Four classes are defined based on the VOTE scheme. V refers to Velum (palate), including soft palate, uvula, lateral velopharyn geal walls. O is for Oropharyngeal lateral walls which include palatine tonsils. T is for Tongue, including tongue base and air way posterior to the tongue base. And, E is for Epiglottis. [4, Page 3469]: Local acoustic characteristics of snore sounds in time frequency spectrogram is learned in convolution layers, [4.1, Page 3469]: The snore sounds are characterized by their sound intensity by a set of spectral parameters [15]. And, we generate a spectrogram for each snore sound clip. A spectrogram is a visual representation of the frequency spectrum over time, and the spectrograms of most sounds have several distinguishing features. In addition, the spectrogram contains frequency and amplitude information over time, and it is important acoustic information to distinguish different types of sounds. Before we compute spectrogram, we upsample the data set as some classes in the dataset has very few samples. We add the small random noise to the sample of small class to obtain more samples. Then, we compute the log power spectrogram using 40 ms frames with a 20 ms shift. The input into the neural networks was normalized to remove any variance. [BRI: High-freq + mouth closed → velum obstruction, Mid-freq + mouth open → oropharyngeal obstruction, Low-freq muffled + mouth closed/mixed → tongue obstruction, Low-freq irregular + mouth closed → epiglottic obstruction] Regarding claim 8 (Previously Presented) Nwe does not explicitly disclose: - other snoring or patient data associated with the subject via an input interface and, in the training mode and/or in the identification mode, take them into account as parameters or parameter signals in classifying the obstruction type. Kopelman discloses: - other snoring or patient data associated with the subject via an input interface and, in the training mode and/or in the identification mode, take them into account as parameters or parameter signals in classifying the obstruction type. [0072]: Systems, methods, devices and apparatus of the present disclosure can comprise one or more sensors adapted to monitor one or more patient parameters, such as physiological data from the patient. The physiological data can be related to one or more of the patient's sleep patterns, sleep apnea events, normal physiological events, and/or abnormal physiological events. In various embodiments, the sensors monitor physiological data from a patient in real-time, and can transmit this sensor data as one or more signals to be received by one or more processors (e.g., of a suitable sleep apnea monitoring and treatment system) It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Regarding claim 9 (Previously Presented) Nwe does not explicitly disclose: - the snoring or patient data comprise at least one of the group consisting of : body mass index, apnea hypopnea index, size of tonsils, size of tongue, Friedman score, time of snoring, duration of sleep. However, Kopelman discloses: - the snoring or patient data comprise at least one of the group consisting of : body mass index, apnea hypopnea index, size of tonsils, size of tongue, Friedman score, time of snoring, duration of sleep. [0099]: The various types of data collected throughout course of the mandibular advancement treatment can be stored (e.g., on one or memory devices associated with the treatment system) for additional processing and analysis. Such data can include data of the patient's previous sleep patterns (e.g., duration of sleep, physiological parameters during sleep), previous sleep apnea patterns (e.g., number, duration, and/or severity of sleep apnea events, symptoms of sleep apnea events, physiological parameters during sleep apnea events) Regarding claim 10 (Previously Presented) Kopelman discloses: - The classification system according to Claim1, the first classifier being at least one machine learning techniques selected from the group consisting of: Support Vector Machine- , Naive- Bayes-System, Least Mean Square method, k-Nearest Neighbours method - k-NN -, Linear Discriminant Analysis - LDA -, Random Forests method - RF -, Extreme Learning Machine - ELM -, Multilayer Perceptron - MLP -, Deep Neural Network - DNN -, logistic regression . [0113]: Machine learning algorithms described herein can comprise support vector machines (SVMs). [0108]: Machine learning algorithms described herein can comprise classification methods, including but not limited to nearest neighbors classifications, or k-nearest neighbors classifications. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Regarding claim 11 (Previously Presented) Nwe does not explicitly disclose: - The classification system according to Claim1, the second classifier being based on one of the following methods of machine learning: Support Vector Machine- , Naive- Bayes-System, Least Mean Square method, k-Nearest Neighbours method - k-NN -, Linear Discriminant Analysis - LDA -, Random Forests method - RF -, Extreme Learning Machine - ELM -, Multilayer Perceptron - MLP -, Deep Neural Network - DNN -, logistic regression However, Kopelman discloses: - The classification system according to Claim1, the second classifier being based on one of the following methods of machine learning: Support Vector Machine- , Naive- Bayes-System, Least Mean Square method, k-Nearest Neighbours method - k-NN -, Linear Discriminant Analysis - LDA -, Random Forests method - RF -, Extreme Learning Machine - ELM -, Multilayer Perceptron - MLP -, Deep Neural Network - DNN -, logistic regression . [0113]: Machine learning algorithms described herein can comprise support vector machines (SVMs). [0108]: Machine learning algorithms described herein can comprise classification methods, including but not limited to nearest neighbors classifications, or k-nearest neighbors classifications. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Regarding claim 12 (Previously Presented) Nwe does not explicitly disclose: - The classification system according to Claim1, the third classifier being based on one of the following methods of machine learning: Support Vector Machine- , Naive- Bayes-System, Least Mean Square method, k-Nearest Neighbours method - k-NN -, Linear Discriminant Analysis - LDA -, Random Forests method - RF -, Extreme Learning Machine - ELM -, Multilayer Perceptron - MLP -, Deep Neural Network - DNN -, logistic regression . However, Kopelman discloses: - The classification system according to Claim1, the third classifier being based on one of the following methods of machine learning: Support Vector Machine- , Naive- Bayes-System, Least Mean Square method, k-Nearest Neighbours method - k-NN -, Linear Discriminant Analysis - LDA -, Random Forests method - RF -, Extreme Learning Machine - ELM -, Multilayer Perceptron - MLP -, Deep Neural Network - DNN -, logistic regression . [0113]: Machine learning algorithms described herein can comprise support vector machines (SVMs). [0108]: Machine learning algorithms described herein can comprise classification methods, including but not limited to nearest neighbors classifications, or k-nearest neighbors classifications. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Regarding claim 21 (Previously Presented) Nwe discloses: - wherein in the third identification mode, the respective obstruction type is identified by the third classifier with an indication of a corresponding probability. [1, Page 3467]: A study which classifies snoring sound based on excitation location is classification between two types of snore sounds: oral and nasal snoring sounds [8]. [2.2, GMM-supervector with Bhattacharyya based kernel, Page 3468]: Bhattacharyya distance is a separability measure between two Gaussian distributions as in equation (2). In equation (2) Σ i a and Σ i b are the covariance matrices of the sounds adapted from Universal Background Model (UBM). And, m a i and m b i   are the adapted mean vectors. Σ i u is the covariance matrix of the UBM. p a and p b   are the probabilistic models,   G M M a and G M M b ,respectively. Regarding claim 28 (Previously Presented) Nwe discloses: - wherein the predefined obstruction types do not include any of the predefined mouth positions. [5, Page 3469]: We use MUNICH-PASSAU SNORE SOUND CORPUS (MPSSC) which is provided by ComParE2017 challenge for snore sound classification based on their excitation location. Four classes are defined based on the VOTE scheme. V refers to Velum (palate), including soft palate, uvula, lateral velopharygeal walls. O is for Oropharyngeal lateral walls which include palatine tonsils. T is for Tongue, including tongue base and air way posterior to the tongue base. And, E is for Epiglottis. The database contains audio samples of 843 snore events from 224 subjects. Please refer to [20] for details of the database Regarding claim 29 (Previously Presented) Nwe discloses: wherein one of the predefined obstruction types specify vibration at a location that is not included in the predefined locations specified by the predefined types of snoring-noise origins. [1, Page 3467]: Snoring occurs when there is a narrowing or obstruction in the upper airways, the turbulence of air and the vibrations of soft palate [1]. Characteristics of the snore sound change depending on the intensity of the vibration on different locations (or excitation location) within the upper airway. [BRI: Perhaps known to a POSITA that current research indicates that snoring is primarily caused by partial obstruction and turbulence in the upper airway, with vibrations of soft tissues (including the soft palate) contributing to the sound]. Regarding claim 30 (Previously Presented) Nwe discloses: - wherein the operations further comprise, at a first time when the subject is asleep and snoring: the generating, [2.1, Page 3468]: Generation of snoring sound is very similar to speech pro duction mechanism [10]. Snore sound is produced by vibration of soft tissues in the upper airway, which acts as an acoustic filter during snoring sound production [1]. As for the speech signal, it is generated by vibrations of the vocal cords, which is filtered and transmitted through the upper airway and oral cav ity [5]. Considering the similarity in production mechanism of speech and snoring sound, it is highly likely that there is acous tic similarity between the two sounds. Hence, we use RT-07 dataset [11] which includes speech, breath, noise, etc.,. to train UBM. [BRI: The first time you snore while asleep is simply the moment your upper airway becomes partially blocked during sleep, causing vibrating soft tissues to produce sound that is often linked to muscle relaxation, nasal congestion, or other airway narrowing factors - the generating, using the first classifier, the snoring-noise origin classification. [4.2, Page 3469]: we employ CNN for snore sound classification. Sounds have global spectral pat terns, as well as local properties such as being more transient or smoother in the time-frequency domain [14]. These can be exposed by using a model that exploits locality leading us to explore two different feature extraction strategies in the context of Convolutional neural networks (CNN) based deep learning. Hence, we employ CNN, a state of the art 2D feature extraction model, to exploit local structures, with log power spectrogram as input in our 3rd sub-system for snore sound classification. [BRI: the 3 rd sub-system is the first classifier] - and the generating, using the third classifier, the obstruction type classification [Introduction, Page 3467]: A study which classifies snoring sound based on excitation location is classification between two types of snore sounds: oral and nasal snoring sounds [8]. This study uses fundamental frequency and the maximum of the amplitude spectrum in the specific band to classify two types of snoring sounds. However, there is still a shortage of robust methods to automatically detect snoring sounds based on their excitation location. The treatment will be effective if it is targeting the area in the upper airways where the snoring sound is generated. In this paper, we propose 3 sub-systems for classifying 4 types of snore sounds based on their excitation location. We then fuse these 3 sub-systems to generate final classification results. Nwe does not explicitly disclose: - the generating, using the second classifier, the mouth position classification. However, Senny discloses: - the generating, using the second classifier, the mouth position classification; [Abstract, Page 87]: given the importance of the detection and classification of sleep apneas and hypopneas (SAHs) in the diagnosis and the characterization of the SAH syndrome, there is a need for a reliable noninvasive technique measuring respiratory effort. This paper proposes a new method for the scoring of SAHs based on the recording of the midsagittal jaw motion (MJM, mouth opening) and on a dedicated automatic analysis of this signal [I, Introduction, Page 88]: The topic of this paper is the automatic detection and classification of SAHs through the midsagittal jaw motion (MJM), a novel Pes surrogate. The rationale is based on the principle that when the upper airway collapses and when, in the meantime, the pressure in the lungs becomes more and more subatmospheric, the jaw is passively lowered by thoracic attraction called thoracic tugs [27]. The consequent opening of the mouth reflects, during sleep, a stronger effort to breath. [I, Introduction, Page 88]: Fig. 1 depicts eight patterns of the MJM signal in cases of (a) OA, (b) CA, (c) MA, (d) H, (e) healthy sleep, (f) snoring, (g) awake and (h) isolated artefacts (swallowing, etc.). To sum up the previous findings [27]–[31], the three main features of the MJM signal to suspect and score SAHs are as follows: 1) a significant opening of the mouth, 2) an SJM occurring after the sleep event (pointed out by an arrow in Fig. 1), and 3) the oscillating jaw movements related to respiratory effort which are expected to be present during an OA. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA. Senny teaches the second classifier (mouth position). One of ordinary skills would be motivated to combine Nwe, Kopelman and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Regarding claim 31(Previously Presented) Nwe discloses: - the generating, using the first classifier, the snoring-noise origin classification; [Abstract, Page 3467]: Snoring is caused by the narrowing of the upper airway and it is excited by different locations within the upper airways. This irregularity could lead to the presence of Obstructive Sleep Apnea Syndrome (OSAS). Diagnosis of OSAS could therefore be made by snoring sound analysis. [4.2, Page 3469]: we employ CNN for snore sound classification. Sounds have global spectral pat terns, as well as local properties such as being more transient or smoother in the time-frequency domain [14]. These can be exposed by using a model that exploits locality leading us to explore two different feature extraction strategies in the context of Convolutional neural networks (CNN) based deep learning. Hence, we employ CNN, a state of the art 2D feature extraction model, to exploit local structures, with log power spectrogram as input in our 3rd sub-system for snore sound classification. [BRI: the 3 rd sub-system is the first classifier] - and the generating, using the third classifier, the obstruction type classification. [Introduction, Page 3467]: A study which classifies snoring sound based on excitation location is classification between two types of snore sounds: oral and nasal snoring sounds [8]. This study uses fundamental frequency and the maximum of the amplitude spectrum in the specific band to classify two types of snoring sounds. However, there is still a shortage of robust methods to automatically detect snoring sounds based on their excitation location. The treatment will be effective if it is targeting the area in the upper airways where the snoring sound is generated. In this paper, we propose 3 sub-systems for classifying 4 types of snore sounds based on their excitation location. We then fuse these 3 sub-systems to generate final classification results. Nwe does not explicitly disclose: - wherein the operations further comprise, at a second time when the subject is asleep, is not snoring, and is experiencing an obstruction related to a snore at a first time However, Kopelman discloses: - wherein the operations further comprise, at a second time when the subject is asleep, is not snoring, and is experiencing an obstruction related to a snore at a first time: [0063]: In some embodiments, the appliance may be activatable prior to snoring when the system identifies patient data or parameters that indicate that snoring or other apnea event is about to begin In [0072]: The sensor data can be indicative of events and/or patient symptoms, such as symptoms associated with the onset of a sleep apnea event, and/or a lessening of symptoms associated with a sleep apnea event. - wherein the operations further comprise, at a second time when the subject is asleep, is not snoring, and is experiencing an obstruction related to a snore at a first time: [0063]: In some embodiments, the appliance may be activatable prior to snoring when the system identifies patient data or parameters that indicate that snoring or other apnea event is about to begin [0072]: The sensor data can be indicative of events and/or patient symptoms, such as symptoms associated with the onset of a sleep apnea event, and/or a lessening of symptoms associated with a sleep apnea event. Nwe and Kopelman do not explicitly disclose: - and the generating, using the second classifier, the mouth position classification; However, Senny discloses: - and the generating, using the second classifier, the mouth position classification; [1, Page 87]: A sleep apnea occurs when breathing stops while sleeping. The clinical definition of an apnea is a decrease in magnitude of the respiration movements lower than 20% of the reference value for at least 10 s. An apnea is either obstructive (OA), central (CA), or mixed (MA) depending on whether respiratory effort is completely present, absent, or partially present during the apnea. [Abstract, Page 87]: given the importance of the detection and classification of sleep apneas and hypopneas (SAHs) in the diagnosis and the characterization of the SAH syndrome, there is a need for a reliable noninvasive technique measuring respiratory effort. This paper proposes a new method for the scoring of SAHs based on the recording of the midsagittal jaw motion (MJM, mouth opening) and on a dedicated automatic analysis of this signal [I, Introduction, Page 88]: The topic of this paper is the automatic detection and classification of SAHs through the midsagittal jaw motion (MJM), a novel Pes surrogate. The rationale is based on the principle that when the upper airway collapses and when, in the meantime, the pressure in the lungs becomes more and more subatmospheric, the jaw is passively lowered by thoracic attraction called thoracic tugs [27]. The consequent opening of the mouth reflects, during sleep, a stronger effort to breath. [I, Introduction, Page 88]: Fig. 1 depicts eight patterns of the MJM signal in cases of (a) OA, (b) CA, (c) MA, (d) H, (e) healthy sleep, (f) snoring, (g) awake and (h) isolated artefacts (swallowing, etc.). To sum up the previous findings [27]–[31], the three main features of the MJM signal to suspect and score SAHs are as follows: 1) a significant opening of the mouth, 2) an SJM occurring after the sleep event (pointed out by an arrow in Fig. 1), and 3) the oscillating jaw movements related to respiratory effort which are expected to be present during an OA. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA. Senny teaches the second classifier (mouth position). One of ordinary skills would be motivated to combine Nwe, Kopelman and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Regarding claim 34 (Previously Presented) Nwe discloses: - wherein at least some of the predefined locations comprise: soft palate, respiratory tract in velopharynx, respiratory tract in the oropharynx, tongue level, and epiglottis level. [5, Page 3469]: We use MUNICH-PASSAU SNORE SOUND CORPUS (MPSSC) which is provided by ComParE2017 challenge for snore sound classification based on their excitation location. Four classes are defined based on the VOTE scheme. V refers to Velum (palate), including soft palate, uvula, lateral velopharyngeal walls. O is for Oropharyngeal lateral walls which include palatine tonsils. T is for Tongue, including tongue base and air way posterior to the tongue base. And, E is for Epiglottis. The database contains audio samples of 843 snore events from 224 subjects [BRI: VOTE is a predefined location] Claims 14, 22 and 24-25 are rejected under 35 U.S.C. 103 as being unpatentable over Tin Lay Nwe et.al (hereinafter Nwe) An Integrated Solution for Snoring Sound Classification Using Bhattacharyya Distance based GMM Supervectors with SVM, Feature Selection with Random Forest and Spectrogram with CNN, INTERSPEECH 2017, August 20–24, 2017, in view of Frédéric Senny et.al (hereinafter Senny) Midsagittal Jaw Movement Analysis for the Scoring of Sleep Apneas and Hypopneas, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 1, JANUARY 2008. Regarding claim 14 (Previously Presented) Nwe discloses: - A method, comprising: [Abstract, Page 3467]: Snoring is caused by the narrowing of the upper airway and it is excited by different locations within the upper airways. This irregularity could lead to the presence of Obstructive Sleep Apnea Syndrome (OSAS). Diagnosis of OSAS could therefore be made by snoring sound analysis. This paper proposes the novel method to automatically classify snoring sounds by their excitation locations for ComParE2017 challenge. - A) training a first classifier with a first plurality of snoring-noise signals with a corresponding type of snoring-noise, wherein the first classifier is adapted to, in a first identification mode, generate a snoring-noise origin classification for a particular snoring-noise signal from a group of predefined types of snoring-noise origin, wherein each of the predefined types of snoring-noise origins specifies a predefined location in a subject's head; [Abstract, Page 3467]: Snoring is caused by the narrowing of the upper airway and it is excited by different locations within the upper airways. This irregularity could lead to the presence of Obstructive Sleep Apnea Syndrome (OSAS). Diagnosis of OSAS could therefore be made by snoring sound analysis. [4.2, Page 3469]: we employ CNN for snore sound classification. Sounds have global spectral patterns, as well as local properties such as being more transient or smoother in the time-frequency domain [14]. These can be exposed by using a model that exploits locality leading us to explore two different feature extraction strategies in the context of Convolutional neural networks (CNN) based deep learning. Hence, we employ CNN, a state of the art 2D feature extraction model, to exploit local structures, with log power spectrogram as input in our 3rd sub-system for snore sound classification. [BRI: the 3 rd sub-system is the first classifier] [5, Page 3469]: We use MUNICH-PASSAU SNORE SOUND CORPUS (MPSSC) which is provided by ComParE2017 challenge for snore sound classification based on their excitation location. Four classes are defined based on the VOTE scheme. V refers to Velum (palate), including soft palate, uvula, lateral velopharyngeal walls. O is for Oropharyngeal lateral walls which include palatine tonsils. T is for Tongue, including tongue base and air way posterior to the tongue base. And, E is for Epiglottis. [BRI: “VOTE” is a standardized classification system that defines four anatomical regions of the upper airway, used to identify the origin (site) of snoring and obstructive sleep apnea, and is applied in both endoscopic evaluation and snoring sound analysis] [2, Page 3468]: Four types of snore sounds considered correspond four different excitation locations within upper airways. [1, Page 3467]: Snoring occurs when there is a narrowing or obstruction in the upper airways, the turbulence of air and the vibrations of soft palate [1]. Characteristics of the snore sound change depending on the intensity of the vibration on different locations (or excitation location) within the upper airway. [BRI: in the VOTE snore-source classification scheme, each predefined type is explicitly tied to a specific anatomical region of the upper airway. These are not arbitrary acoustic classes but they correspond to distinct, predefined excitation locations in the subject’s head where the vibration or obstruction producing the snore originates. Snoring that results from narrowing or obstruction in the upper airway can indeed originate from a specific, predetermined anatomical site in the head and neck, but the exact location varies between individuals and can change during sleep] C) training a third classifier with a third plurality of training data comprising i) particular snoring-noise origins, and ii) particular mouth positions, wherein the third classifier is adapted to, in a third identification mode, generate an obstruction type classification from a group of predefined obstruction types of obstructive sleep apnea, wherein at least some of the predefined obstruction types specify constriction at one of the predefined locations. [2.2, Page 3468]: As for the speech signal, it is generated by vibrations of the vocal cords, which is filtered and transmitted through the upper airway and oral cavity [5]. [Introduction, Page 3467]: A study which classifies snoring sound based on excitation location is classification between two types of snore sounds: oral and nasal snoring sounds [8]. This study uses fundamental frequency and the maximum of the amplitude spectrum in the specific band to classify two types of snoring sounds. However, there is still a shortage of robust methods to automatically detect snoring sounds based on their excitation location. The treatment will be effective if it is targeting the area in the upper airways where the snoring sound is generated. In this paper, we propose 3 sub-systems for classifying 4 types of snore sounds based on their excitation location. We then fuse these 3 sub-systems to generate final classification results. [1, Page 3467]: We employ machine learning methods such as Support Vector Machine(SVM), Random Forest (RF) and Convolutional Neural Network(CNN) as classifiers for the three sub-systems respectively. System block diagram of our snore sound classification is shown in Figure 1. PNG media_image1.png 268 643 media_image1.png Greyscale [6, Page 3469]: We conduct several experiments to investigate the effectiveness of the individual sub-systems and fused system. Firstly, we examine effect of integrating Bhattacharyya based GMM super vector to the baseline feature set. We use Support Vector Ma chine(SVM) for classification with the integrated feature set. Secondly, we investigate feature selection method, CFS, to improve the performance of the baseline system by using a subset of baseline feature set which is important for classification of 4 different snore sounds, [6.1, Page 3470]: We use training set to train the snore sound models. The result in Table 1 shows that snore sound classification accuracy improves 7.7% absolute over the challenge baseline system for development set (3rd row of Table 1) when we integrate Bhattacharyya based GMM supervectors (Bhat-GMM-Sup) to baseline features, PNG media_image2.png 283 639 media_image2.png Greyscale Nwe does not explicitly disclose: - B) training a second classifier with a second plurality of snoring-noise signals with a corresponding mouth position, wherein the second classifier is adapted to, in a second identification mode, generate a mouth position classification for the particular snoring-noise signal from a group of predefined types of mouth positions; However, Senny discloses: - B) training a second classifier with a second plurality of snoring-noise signals with a corresponding mouth position, wherein the second classifier is adapted to, in a second identification mode, generate a mouth position classification for the particular snoring-noise signal from a group of predefined types of mouth positions; [1, Page 87]: A sleep apnea occurs when breathing stops while sleeping. The clinical definition of an apnea is a decrease in magnitude of the respiration movements lower than 20% of the reference value for at least 10 s. An apnea is either obstructive (OA), central (CA), or mixed (MA) depending on whether respiratory effort is completely present, absent, or partially present during the apnea. [Abstract, Page 87]: given the importance of the detection and classification of sleep apneas and hypopneas (SAHs) in the diagnosis and the characterization of the SAH syndrome, there is a need for a reliable noninvasive technique measuring respiratory effort. This paper proposes a new method for the scoring of SAHs based on the recording of the midsagittal jaw motion (MJM, mouth opening) and on a dedicated automatic analysis of this signal [I, Introduction, Page 88]: The topic of this paper is the automatic detection and classification of SAHs through the midsagittal jaw motion (MJM), a novel Pes surrogate. The rationale is based on the principle that when the upper airway collapses and when, in the meantime, the pressure in the lungs becomes more and more subatmospheric, the jaw is passively lowered by thoracic attraction called thoracic tugs [27]. The consequent opening of the mouth reflects, during sleep, a stronger effort to breath. [I, Introduction, Page 88]: Fig. 1 depicts eight patterns of the MJM signal in cases of (a) OA, (b) CA, (c) MA, (d) H, (e) healthy sleep, (f) snoring, (g) awake and (h) isolated artefacts (swallowing, etc.). To sum up the previous findings [27]–[31], the three main features of the MJM signal to suspect and score SAHs are as follows: 1) a significant opening of the mouth, 2) an SJM occurring after the sleep event (pointed out by an arrow in Fig. 1), and 3) the oscillating jaw movements related to respiratory effort which are expected to be present during an OA. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA. One of ordinary skills would be motivated to combine Nwe, Kopelman and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Regarding claim 22 (Previously Presented) Nwe discloses: - wherein the group of predefines types of snoring-noise origin comprise velopharynx, oropharynx, tongue base area and/or epiglottis area. [5, Page 3469]: We use MUNICH-PASSAU SNORE SOUND CORPUS (MPSSC) which is provided by ComParE2017 challenge for snore sound classification based on their excitation location. Four classes are defined based on the VOTE scheme. V refers to Velum (palate), including soft palate, uvula, lateral velopharyngeal walls. O is for Oropharyngeal lateral walls which include palatine tonsils. T is for Tongue, including tongue base and air way posterior to the tongue base. And, E is for Epiglottis. The database contains audio samples of 843 snore events from 224 subjects. Regarding claim 24 (Previously Presented) Nwe does not explicitly disclose: - wherein the group of predefined types of mouth positions comprise mouth open, mouth closed. However, Senny discloses: [II, Page 88]: The mandibular movement signal can be expressed either in absolute value (millimeter) or in normalized value (percentage of mouth opening). Normalizing the signal is a two-steps process. First, patient is asked to widely open then fully close his/her mouth just at the very beginning of the night recording. Second, the user must choose the two values mouth” and vc “wide open “fully closed mouth” through a graphical user interface displaying the first minutes of the signal. Data are then normalized PNG media_image3.png 63 587 media_image3.png Greyscale The 100% corresponds to a wide open mouth and 0% to a fully closed mouth. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). One of ordinary skills would be motivated to combine Nwe , and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Regarding claim 25 (Previously Presented) Nwe does not explicitly disclose: - wherein the group of predefined types of mouth positions comprise: mouth open, mouth closed, and intermediate mouth positions. However, Senny discloses: [II, Page 88]: The mandibular movement signal can be expressed either in absolute value (millimeter) or in normalized value (percentage of mouth opening). Normalizing the signal is a two-steps process. First, patient is asked to widely open then fully close his/her mouth just at the very beginning of the night recording. Second, the user must choose the two values mouth” and vc “wide open “fully closed mouth” through a graphical user interface displaying the first minutes of the signal. Data are then normalized PNG media_image3.png 63 587 media_image3.png Greyscale The 100% corresponds to a wide open mouth and 0% to a fully closed mouth. In this paper, the mandibular movement signal is displayed in percentage of mouth opening for convenience, but analyzed in millimeter to avoid user’s measure errors occurring in the normalization step [BRI: within the context of percentage, any percentage between above 0 and below 100 relatively may indicate an intermediate mouth position] It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, and Senny. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). One of ordinary skills would be motivated to combine Nwe , and Senny that can improve the classification of SAHs by making use of respiratory events with the jaw motion recognized as a reliable signal to score SAHs and may be a very useful physiological signal to screen SDB (SAH, RERA, or snoring) (Senny [IV, Page 94]). Claims 15-18 are rejected under 35 U.S.C. 103 as being unpatentable over Tin Lay Nwe et.al (hereinafter Nwe) An Integrated Solution for Snoring Sound Classification Using Bhattacharyya Distance based GMM Supervectors with SVM, Feature Selection with Random Forest and Spectrogram with CNN, INTERSPEECH 2017, August 20–24, 2017. in view of Frédéric Senny et.al (hereinafter Senny) Midsagittal Jaw Movement Analysis for the Scoring of Sleep Apneas and Hypopneas, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 1, JANUARY 2008. further in view of Hisham Alshaer et.al (hereinafter Alshaer) US 2014/0188006 A1. Regarding claim 15: (Previously Presented) Nwe and Senny do not explicitly disclose: - wherein the training of the first classifier and the training of the second classifier take place separately from one another, wherein the first classifier is trained and learns independently of the mouth position and the second classifier is trained independently of the type of snoring-noise origin. However, Alshaer discloses: - wherein the training of the first classifier and the training of the second classifier take place separately from one another, wherein the first classifier is trained and learns independently of the mouth position and the second classifier is trained independently of the type of snoring-noise origin. [0195]: An example of a classification model based on this approach is provided in FIG. 26, in accordance with one illustrative embodiment of the invention. In particular, process 2600 can be subdivided into two main branches: a training phase 2602 and an implementation phase 2604. During the training phase 2602, a known data set 2606 (e.g. breath sounds recorded during known apnea/hypopnea events independently associated with OSA) are processed (e.g. via RAPT) so to first isolate periodic breath sound segments and extract therefrom respective pitch contours for known obstructive and non-obstructive snoring segments (steps 2608 and 2610, respectively). Extracted contours are then characterized (e.g. via FDA, as in the below example) at steps 2612 and 2614, and the distinguishable characteristics thereof retained in training a classifier 2616 selected so to produce classification criteria 2618 usable in subsequent classifications. Various pitch contour characteristics in the time, frequency and time/frequency domain may be selected in optimizing classification criteria based on a given training data set, as will be readily appreciated by the skilled artisan. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Senny and Alshaer. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). Alshaer teaches first classifier is trained independently of the mouth position. One of ordinary skills would have been motivated to combine Nwe, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Regarding claim 16 (Original) Nwe discloses: - the training and learning of the first and the second classifier take place with a time shift. [4.1, Page 3469]: The snore sounds are characterized by their sound intensity by a set of spectral parameters [15]. And, we generate a spectrogram for each snore sound clip. A spectrogram is a visual representation of the frequency spectrum over time, and the spectrograms of most sounds have several distinguishing features. In addition, the spectrogram contains frequency and amplitude information over time, and it is important acoustic information to distinguish different types of sounds. Before we compute spectrogram, we up sample the data set as some classes in the dataset has very few samples. We add the small random noise to the sample of small class to obtain more samples. Then, we compute the log power spectrogram using 40 ms frames with a 20 ms shift. The input into the neural networks was normalized to remove any variance [BRI: time process is essentially sequential model training where each classifier is fitted to data available at its own training time. A spectrogram is a time–frequency representation of a signal, showing how the frequency content of a sound changes over time. Different sounds have distinct patterns in the time–frequency domain — e.g., speech has periodic harmonics, music has sustained tones, and environmental sounds may have transient bursts. The log power spectrogram captures these patterns in a compact form that machine learning models can use to distinguish sound types] Yes — jaw movement patterns and the temporomandibular joint (TMJ) classification directly relate to mouth mechanics, and TMJ dysfunction can contribute to snoring through airway changes. Regarding claim 17: (Original) Nwe and Senny do not explicitly disclose: - wherein the training and learning of the first and the second classifier take place simultaneously. However, Alshaer discloses: [0178]: During the training phase 2202, a known data set consisting of known OSA (2206) and CSA (2207) events (e.g. breath sounds recorded during known apnea/hypopnea events independently associated with CSA and OSA, respectively) are processed, [0217]: With reference to FIG. 28, a system 2800, similar to that depicted in FIG. 1, is shown as used to develop and validate a method for UA narrowing detection via breath sound analysis, implemented in accordance with one embodiment of the invention. The system 2800 generally comprises a face mask 2812 having a microphone 2802 embedded therein for disposal at a distance from a nose and mouth area of the subject's face, from which breath sounds may be recorded [0217]: At step 3024, the recorded breath sounds are first processed (e.g. via LPC) so to generate extractable features 3026 (e.g. one or more LPC coefficients) to be compared by classifier 3028 with the preset UA narrowing classification criteria 3020, in classifying the processed event as indicative of an open or narrowed airway. For example, the classifier 3028 may be configured to output a narrowed airway indication where extracted features fall within a preset range or above or below a preset threshold and otherwise default to an open airway indication [BRI: An “open” airway indicator means the airway passage is clear and visible when the mouth is open; a “narrowed” airway indicator means there is obstruction or reduced space in the airway. The mouth position is a prerequisite for the assessment [0047]: The plurality of different lower jaw positions available provides the advancement apparatus with a range of control over the position of the lower jaw of the patient relative to the upper jaw of the patient. For example, the amount of mandibular advancement, the amount of vertical displacement between upper and lower jaws, and/or the amount of opening of the patient's mouth can all be controlled according to a treatment plan and/or a determined effectiveness of one or more lower jaw positions with respect to treatment of sleep apnea. [0197]: Functional Data Analysis (FDA) can be used to provide automatic distinction between regular snores and those associated with obstruction. [BRI: automatic classification systems can distinguish regular snores from those linked to obstruction, and these systems are trained on combined snoring and obstruction-related features, effectively representing a joint training approach] [0062]: FIG. 2 provides another example of a mask 200 usable in acquiring breathing sounds suitable in the present context. In this example, the mask 200 generally comprises at least one transducer, such as microphones 202 and 204, and a support structure 206 for supporting same above a nose and mouth area of the subject's face. The support structure 206 is generally shaped and configured to rest on the subject's face and thereby delineate the nose and mouth area thereof, and comprises two or more outwardly projecting limbs 208 (e.g. three limbs in this example) that, upon positioning the mask 200, converge into a transducer supporting portion 210 for supporting microphones 202 and 204 at a distance from this area. [BRI: the mouth area of a subject’s face can be used as a recording site for breath sounds, and it is feasible to train classifiers for both snore sounds and general mouth-area breath sounds using shared or joint training approaches] It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Senny and Alshaer. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). One of ordinary skills would have been motivated to combine Nwe, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Regarding claim 18: (Previously Presented) Nwe and Senny do not explicitly disclose: - wherein the training and learning of the first and the second classifier take place together and simultaneously. However, Alshaer discloses: - wherein the training and learning of the first and the second classifier take place together and simultaneously. [0178]: During the training phase 2202, a known data set consisting of known OSA (2206) and CSA (2207) events (e.g. breath sounds recorded during known apnea/hypopnea events independently associated with CSA and OSA, respectively) are processed, [0217]: With reference to FIG. 28, a system 2800, similar to that depicted in FIG. 1, is shown as used to develop and validate a method for UA narrowing detection via breath sound analysis, implemented in accordance with one embodiment of the invention. The system 2800 generally comprises a face mask 2812 having a microphone 2802 embedded therein for disposal at a distance from a nose and mouth area of the subject's face, from which breath sounds may be recorded [0047]: The plurality of different lower jaw positions available provides the advancement apparatus with a range of control over the position of the lower jaw of the patient relative to the upper jaw of the patient. For example, the amount of mandibular advancement, the amount of vertical displacement between upper and lower jaws, and/or the amount of opening of the patient's mouth can all be controlled according to a treatment plan and/or a determined effectiveness of one or more lower jaw positions with respect to treatment of sleep apnea. [0245]: With reference to FIG. 33, and in accordance with one embodiment of the invention, the local outputs form the various processes described above generally with reference to FIGS. 6A and 6B, can be combined to produce a global output determination indicative of the most likely characterization of the subject's condition. In this example, the global classifier 3300 receives a local output indication from each of the aperiodic sound evaluation module 3302 (e.g. output from step 662 of FIG. 6A; narrowed airway output 634 or open airway output 636 from step 632 of FIG. 6B); periodic sound evaluation module 3304 (e.g. output from step 660 of FIG. 6A; collapsible airway output 630 or stable airway output 628 from step 626 of FIG. 6B); and sound amplitude profile evaluation module 3306 (e.g. output from step 658 of FIG. 6A; gradual fall/abrupt rise output 618 or crescendo/decrescendo output 616 from step 614 of FIG. 6B). A pre-selected weighing factor is then applied to each local output at step 3308 so to adjust an effect each one of these outputs is to have on the global output. [0245]: a local output may be provided along with a confidence value automatically calculated thereon as a function of a classification accuracy estimated by each processing branch. For example, where a local output was classified based on a comparison of the output value with a threshold value or range, [0245]: a simple majority voting process whereby one of CSA and OSA is deemed to be the most likely classification, or again by way of a weighed sum of respective local outputs to produce an output probability for each possible output, to name a few. Where conflicting local outputs are entered, the system may be configured to output an error or "unclassifiable" code, or again output details as to the various conflicts identified between respective local outputs. In another example, the system may rather be configured to output a default value (e.g. OSA) unless a combination of local outputs exceeds a preset probability threshold (e.g. 75%) for an alternative classification (e.g. CSA). Likewise, the global output indicator 3312 may also be configured to output a severity index or value (e.g. AHI), as shown by output 640 of FIG. 6B It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Senny and Alshaer. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). One of ordinary skills would have been motivated to combine Nwe, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Claims 19-20 and 27 are rejected under 35 U.S.C. 103 as being unpatentable over Tin Lay Nwe et.al (hereinafter Nwe) An Integrated Solution for Snoring Sound Classification Using Bhattacharyya Distance based GMM Supervectors with SVM, Feature Selection with Random Forest and Spectrogram with CNN, INTERSPEECH 2017, August 20–24, 2017. in view of Frédéric Senny et.al (hereinafter Senny) Midsagittal Jaw Movement Analysis for the Scoring of Sleep Apneas and Hypopneas, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 1, JANUARY 2008. further in view of Avi Kopelman et.al. (hereinafter Kopelman) US 2016/0199215 A1. Regarding claim 19: (Previously Presented) Nwe and Senny do not explicitly disclose: - wherein in the first identification mode, the respective types of snoring-noise origin are identified by the first classifier with respective probability values However, Kopelman discloses: - wherein in the first identification mode, the respective types of snoring-noise origin are identified by the first classifier with respective probability values [0005]: A mandibular advancement device can be combined with patient monitoring and customized treatment to treat obstructive sleep apnea and snoring with improved detection of symptoms associated with sleep apnea and improved treatment of sleep apnea based on or in response to a patient's sleep apnea status, [0006]: onset of a sleep apnea event based on the set of sensor data, and transmit a control signal to the intraoral appliance to cause the intraoral appliance to displace a lower jaw of the patient from a first position to a second position in order to treat the sleep apnea event, [0125]: Machine learning algorithms described herein can comprise anomaly detection and/or outlier detection that can be used to identify physiological data that do not conform to an expected pattern or are otherwise distinct from other physiological data in a dataset, Anomaly detection and/or outlier detection can comprise, without limitation, density-based techniques, k-nearest neighbors classification, local outlier factor analysis, subspace-based outlier detection, correlation-based outlier detection, support vector machines, replicator neural networks, cluster analysis, deviations from association rules, deviations from frequent item sets, fuzzy logic based outlier detection, ensemble techniques, feature bagging, score normalization, and/or variants thereof and/or combinations thereof, [0128]: the ensemble learning method comprises random forests that operate by constructing a plurality of decision trees and outputting the class that is the mode of the classes (classification) or mean prediction (regression) of the individual trees. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Senny and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Regarding claim 20: (Previously Presented) Nwe and Senny do not explicitly disclose: - the respective mouth position are identified by the second classifier with respective probability values and fed to the third classifier for identification of the obstruction type. However, Kopelman discloses: - the respective mouth position are identified by the second classifier with respective probability values and fed to the third classifier for identification of the obstruction type. [0005]: A mandibular advancement device can be combined with patient monitoring and customized treatment to treat obstructive sleep apnea and snoring with improved detection of symptoms associated with sleep apnea and improved treatment of sleep apnea based on or in response to a patient's sleep apnea status, [0006]: onset of a sleep apnea event based on the set of sensor data, and transmit a control signal to the intraoral appliance to cause the intraoral appliance to displace a lower jaw of the patient from a first position to a second position in order to treat the sleep apnea event, [0125]: Machine learning algorithms described herein can comprise anomaly detection and/or outlier detection that can be used to identify physiological data that do not conform to an expected pattern or are otherwise distinct from other physiological data in a dataset, Anomaly detection and/or outlier detection can comprise, without limitation, density-based techniques, k-nearest neighbors classification, local outlier factor analysis, subspace-based outlier detection, correlation-based outlier detection, support vector machines, replicator neural networks, cluster analysis, deviations from association rules, deviations from frequent item sets, fuzzy logic based outlier detection, ensemble techniques, feature bagging, score normalization, and/or variants thereof and/or combinations thereof, [0128]: the ensemble learning method comprises random forests that operate by constructing a plurality of decision trees and outputting the class that is the mode of the classes (classification) or mean prediction (regression) of the individual trees, It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Senny and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe, Senny and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Regarding claim 27 (Previously Presented) Nwe and Senny do not explicitly disclose - the snoring or patient data comprise at least one of the group consisting of : body mass index, apnea hypopnea index, size of tonsils, size of tongue, Friedman score, time of snoring, duration of sleep. However, Kopelman discloses: - the snoring or patient data comprise at least one of the group consisting of : body mass index, apnea hypopnea index, size of tonsils, size of tongue, Friedman score, time of snoring, duration of sleep. [0099]: The various types of data collected throughout course of the mandibular advancement treatment can be stored (e.g., on one or memory devices associated with the treatment system) for additional processing and analysis. Such data can include data of the patient's previous sleep patterns (e.g., duration of sleep, physiological parameters during sleep), previous sleep apnea patterns (e.g., number, duration, and/or severity of sleep apnea events, symptoms of sleep apnea events, physiological parameters during sleep apnea events), or previous mandibular advancement treatments applied. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Senny and Kopelman. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Senny teaches the second classifier (mouth position). Kopelman teaches probability values. One of ordinary skills would be motivated to combine Nwe and Kopelman that can use machine learning algorithm to determine a modified position of the lower jaw to improve the effectiveness of treating the sleep apnea event; and transmit a third control signal to the intraoral appliance to displace the lower jaw to the modified position (Kopelman [0026]). Claims 2, 5, 7, and 32 are rejected under 35 U.S.C. 103 as being unpatentable over Tin Lay Nwe et.al (hereinafter Nwe) An Integrated Solution for Snoring Sound Classification Using Bhattacharyya Distance based GMM Supervectors with SVM, Feature Selection with Random Forest and Spectrogram with CNN, INTERSPEECH 2017, August 20–24, 2017, In view of Avi Kopelman et.al. (hereinafter Kopelman) US 2016/0199215 A1, in view of Frédéric Senny et.al (hereinafter Senny) Midsagittal Jaw Movement Analysis for the Scoring of Sleep Apneas and Hypopneas, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 1, JANUARY 2008. further in view of Hisham Alshaer et.al (hereinafter Alshaer) US 2014/0188006 A1. Regarding claim 2: (Previously Presented) Nwe, Kopelman, Senny do not disclose: - the first classifier and the second classifier being adapted such that the respective training of the first and of the second classifier with a plurality of snoring-noise signals can be performed separately from one another, with the first classifier training and learning independently of the mouth position and the second classifier independently of the type of snoring- noise origin However, Alshaer discloses: [0195]: An example of a classification model based on this approach is provided in FIG. 26, in accordance with one illustrative embodiment of the invention. In particular, process 2600 can be subdivided into two main branches: a training phase 2602 and an implementation phase 2604. During the training phase 2602, a known data set 2606 (e.g. breath sounds recorded during known apnea/hypopnea events independently associated with OSA) are processed (e.g. via RAPT) so to first isolate periodic breath sound segments and extract therefrom respective pitch contours for known obstructive and non-obstructive snoring segments (steps 2608 and 2610, respectively). Extracted contours are then characterized (e.g. via FDA, as in the below example) at steps 2612 and 2614, and the distinguishable characteristics thereof retained in training a classifier 2616 selected so to produce classification criteria 2618 usable in subsequent classifications. Various pitch contour characteristics in the time, frequency and time/frequency domain may be selected in optimizing classification criteria based on a given training data set, as will be readily appreciated by the skilled artisan. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman, Senny and Alshaer. Nwe teaches first classifier and third classifier. Kopelman teaches systems associated with OSA. Senny teaches second classifier. Alshaer teaches independently of the mouth position and the second classifier independently of the type of snoring- noise origin One of ordinary skills would have been motivated to combine Nwe, Kopelman, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Regarding claim 5: (Previously Presented) Nwe, Kopelman, and Senny do not explicitly disclose: - the second classifier being adapted to identify, indicate and forward to the third classifier in the identification mode, the respective mouth position with a respective probability However, Alshaer discloses: - the second classifier being adapted to identify, indicate and forward to the third classifier in the identification mode, the respective mouth position with a respective probability [0217]: With reference to FIG. 28, a system 2800, similar to that depicted in FIG. 1, is shown as used to develop and validate a method for UA narrowing detection via breath sound analysis, implemented in accordance with one embodiment of the invention. The system 2800 generally comprises a face mask 2812 having a microphone 2802 embedded therein for disposal at a distance from a nose and mouth area of the subject's face, from which breath sounds may be recorded, for example as shown illustratively by sample waveform 2830. Face masks as shown in the embodiments of FIGS. 2 to 4, and others like them, may also be used in this context, as will be understood by the skilled artisan. Pharyngeal catheters 2840 and a pneumotachometer 2850, as used in the below-described example, are also shown for purpose of validating breath sound analysis, and in generating a training data set from which classification criteria may be identified and set for the subsequent automated classification of unknown data sets. A recording/processing module (not shown), such as recording/processing module 120, 220 and 330 of FIGS. 1, 2, and 3, respectively, is again included to record breath sounds captured by the microphone 2802, and process same in implementing, at least in part, the steps described below. [0076]: In the embodiments of FIGS. 1, 3 and 4, however, a single microphone may alternatively be used to capture both sound and airflow, wherein each signal may be optionally distinguished and at least partially isolated via one or more signal processing techniques, [0062]: FIG. 2 provides another example of a mask 200 usable in acquiring breathing sounds suitable in the present context. [0062]: The support structure 206 is generally shaped and configured to rest on the subject's face and thereby delineate the nose and mouth area thereof [0066]: the support structure 306 is shaped and configured to support the transducer 302 above the nose and mouth area at a preset orientation in relation thereto, wherein the preset orientation may comprise one or more of a preset position and a preset angle to intercept airflow produced by both the subject's nose and mouth [0060]: As schematically depicted, the one or more transducers 102 are operatively coupled to a data recording/processing module 120 for recording breath sound data illustratively in In [6.2, Page 115]: Performance of feature subsets Snoring and speech have a lot of acoustic similarities: both are generated in the upper airway through vibrations caused by airflow, acoustically shaped by the frequency transfer function of the upper airway and emitted through mouth and nose. It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman, Senny and Alshaer. Nwe teaches first classifier and third classifier. Kopelman teaches systems associated with OSA. Senny teaches second classifier. Alshaer teaches independently of the mouth position and the second classifier independently of the type of snoring- noise origin One of ordinary skills would have been motivated to combine Nwe, Kopelman, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Regarding claim 7: (Previously Presented) Nwe , Kopelman, and Senny do not explicitly disclose: - indicate and forward to an output interface to the respective obstruction type with a respective probability for display However, Alshaer discloses: - indicate and forward to an output interface to the respective obstruction type with a respective probability for display [0191]: In one exemplary embodiment, periodicity of the recorded sound is identified via a Robust Algorithm for Pitch Tracking (RAPT), which can be used not only to distinguish periodic from aperiodic sounds, but also calculate the pitch of periodic sounds, which calculated pitch can then be used for pitch stability analysis. As will be appreciated by the skilled artisan, RAPT has traditionally been used for detecting the fundamental frequency or pitch in speech analysis. [0191]: the RAPT process is generally configured to output for each processed window a periodicity identifier (e.g. 1 for periodic and 0 for aperiodic), and where periodicity is identified, a pitch frequency and probability or accuracy measure (e.g. based on signal autocorrelation), as well as other outputs not currently being used in current implementations. [0020]: In accordance with another embodiment, the above methods are automatically implemented by one or more processors of a computing system, and further comprise outputting, via a user interface, an indication of a candidate's condition. [0062]: The support structure 206 is generally shaped and configured to rest on the subject's face and thereby delineate the nose and mouth area thereof [0066]: the support structure 306 is shaped and configured to support the transducer 302 above the nose and mouth area at a preset orientation in relation thereto, wherein the preset orientation may comprise one or more of a preset position and a preset angle to intercept airflow produced by both the subject's nose and mouth [0079]: the processing module may further be coupled to, or operated in conjunction with, an external processing and/or interfacing device, such as a local or remote computing device or platform provided for the further processing and/or display of raw and/or processed data, or again for the interactive display of system implementation data, protocols and/or diagnostics tools. [0083]: The device 500 may further comprise a user interface 530, either integral thereto, or distinctly and/or remotely operated therefrom for the input of data and/or commands (e.g. keyboard, mouse, scroll pad, touch screen, push-buttons, switches, etc.) by an operator thereof, and/or for the presentation of raw, processed and/or diagnostic data with respect to breathing disorder identification, characterization and/or diagnosis (e.g. graphical user interface such as CRT, LCD, LED screen or the like, visual and/or audible signals / alerts / warnings / cues, numerical displays, etc.). It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe , Kopelman, Senny and Alshaer. Nwe teaches first classifier and third classifier. Kopelman teaches systems associated with OSA. Senny teaches second classifier. Alshaer teaches independently of the mouth position and the second classifier independently of the type of snoring- noise origin One of ordinary skills would have been motivated to combine Nwe, Kopelman, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Regarding claim 32 (Previously Presented) Nwe, Kopelman and Senny do not explicitly disclose: wherein the third identification mode comprises a matrix probability calculation using i) a first input vector from the predefined types of snoring-noise origin, and ii) at least one second input vector of the predefined mouth positions. However, Alshaer discloses: wherein the third identification mode comprises a matrix probability calculation using i) a first input vector from the predefined types of snoring-noise origin, [0053]: FIG. 28 is a schematic diagram of a system for validating upper airway (UA) narrowing detection achieved via breath sound analysis in accordance with one embodiment of the invention; [0054]: FIG. 29 is a diagram of an analogy relied upon for UA narrowing detection, in accordance with one embodiment of the invention, between a Linear Prediction Coding (LPC) modeling of unvoiced speech sounds and that of turbulent breath sounds; [0235]: To measure the ability of k-mean to separate LPC vectors in M based on the underlying .sub.UA status, BL and peak R.sub.UA, the sum of LPC vectors in each of the 2 resulting clusters for each status was calculated at step 3012 as: PNG media_image4.png 72 306 media_image4.png Greyscale which is the sum of the LPC vectors x.sub.1 in each inspiratory sound segment s.sub.1, where n is the total number of vectors in M, l is the number of inspiratory segments in the data set, [0217]: With reference to FIG. 28, a system 2800, similar to that depicted in FIG. 1, is shown as used to develop and validate a method for UA narrowing detection via breath sound analysis, implemented in accordance with one embodiment of the invention. The system 2800 generally comprises a face mask 2812 having a microphone 2802 embedded therein for disposal at a distance from a nose and mouth area of the subject's face, from which breath sounds may be recorded, [0098]: A narrowing may be more readily associated with OSA, as opposed to aperiodic sounds indicative of an open UA, In [0209]: Accordingly, a characteristic mean and standard deviation can be generated for each condition (obstructive vs. non-obstructive), against which a test curve or group of curves representative of a new data set (e.g. extracted pitch contour(s) from unclassified periodic breath sound recording(s)) can be compared to yield a statistical result indicating the proximity of the test curve to either of the 2 families, thus providing an indication as to a most probable condition of the tested candidate (i.e. normal or CSA snoring vs. OSA snoring), [0178 ]: During the training phase 2202, a known data set consisting of known OSA (2206) and CSA (2207) events (e.g. breath sounds recorded during known apnea/hypopnea events independently associated with CSA and OSA, respectively) are processed [0183 ]:The objective is to compute the matching cost: DTW(p, q). To align the two sequences using DTW, an n x m matrix is constructed where the (i, j)-th entry of the matrix indicates the distance d(p.sub.i, q.sub.j) between the two points p.sub.i and q.sub.j, where d(p.sub.i, q.sub.i)=(p.sub.i, q.sub.l).sup.2. [0099]: a global output may consist of an overall classification or indication as to the candidate's most likely condition (e.g. OSA or CSA) along with an indication as to a severity of the reported condition (e.g. AHI). In other embodiments, a probability or likelihood may be associated with each condition and ii) at least one second input vector of the predefined mouth positions. [0053]: FIG. 28 is a schematic diagram of a system for validating upper airway (UA) narrowing detection achieved via breath sound analysis in accordance with one embodiment of the invention; [0054]: FIG. 29 is a diagram of an analogy relied upon for UA narrowing detection, in accordance with one embodiment of the invention, between a Linear Prediction Coding (LPC) modeling of unvoiced speech sounds and that of turbulent breath sounds; [0235]: To measure the ability of k-mean to separate LPC vectors in M based on the underlying .sub.UA status, BL and peak R.sub.UA, the sum of LPC vectors in each of the 2 resulting clusters for each status was calculated at step 3012 as: PNG media_image4.png 72 306 media_image4.png Greyscale which is the sum of the LPC vectors x.sub.1 in each inspiratory sound segment s.sub.1, where n is the total number of vectors in M, l is the number of inspiratory segments in the data set, [0217]: With reference to FIG. 28, a system 2800, similar to that depicted in FIG. 1, is shown as used to develop and validate a method for UA narrowing detection via breath sound analysis, implemented in accordance with one embodiment of the invention. The system 2800 generally comprises a face mask 2812 having a microphone 2802 embedded therein for disposal at a distance from a nose and mouth area of the subject's face, from which breath sounds may be recorded, [0098]: A narrowing may be more readily associated with OSA, as opposed to aperiodic sounds indicative of an open UA, [0209]: Accordingly, a characteristic mean and standard deviation can be generated for each condition (obstructive vs. non-obstructive), against which a test curve or group of curves representative of a new data set (e.g. extracted pitch contour(s) from unclassified periodic breath sound recording(s)) can be compared to yield a statistical result indicating the proximity of the test curve to either of the 2 families, thus providing an indication as to a most probable condition of the tested candidate (i.e. normal or CSA snoring vs. OSA snoring). It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman, Senny and Alshaer. Nwe teaches first classifier and third classifier. Kopelman teaches systems associated with OSA. Senny teaches second classifier. Alshaer teaches independently of the mouth position and the second classifier independently of the type of snoring- noise origin One of ordinary skills would have been motivated to combine Nwe, Kopelman, Senny, and Alshaer to present improvements for providing less invasive approach to sleep apnea identification (Alshaer [0059]). Claim 33 is rejected under 35 U.S.C. 103 as being unpatentable over Tin Lay Nwe et.al (hereinafter Nwe) An Integrated Solution for Snoring Sound Classification Using Bhattacharyya Distance based GMM Supervectors with SVM, Feature Selection with Random Forest and Spectrogram with CNN, INTERSPEECH 2017, August 20–24, 2017, In view of Avi Kopelman et.al. (hereinafter Kopelman) US 2016/0199215 A1, in view of Frédéric Senny et.al (hereinafter Senny) Midsagittal Jaw Movement Analysis for the Scoring of Sleep Apneas and Hypopneas, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 1, JANUARY 2008. further in view Manfred MUELLER (hereinafter MUELLER) US 2017/0055841 A1. Regarding claim 33 (Previously Presented) Nwe , Kopelman, Senny and Alshaer do not explicitly disclose: - wherein some of the predefined obstruction types comprise: anterior-posterior constriction, lateral constriction, and circular constriction. [0004]: The Obstructive Sleep Apnea (OSA) patient pool eligible for treatments other than Positive Airway Pressure (PAP) is increasing as nowadays more alternatives become available [0093]: The concept as embodied in the description herein is an optical sensor which may be applied to a flexible or semi-flexible catheter which can be placed into the upper airway of a patient to measure changes of airway geometry during natural sleep [0093]: In some embodiments the number and spacing of the sensors is such that it is possible to interpolate the airway cross section at these key anatomical locations where the key location occurs between sensors. From the change of cross section a data processing unit may be configured to calculate key physiological parameters like the percentage of narrowing and/or configuration of narrowing (for example whether the narrowing is an anterior-posterior narrowing, a lateral narrowing or a circular or radial narrowing). It would have been obvious to one of ordinary skills in the art before the effective filing date of the present application to combine Nwe, Kopelman , Senny and MUELLER. Nwe teaches sleep apnea and first classification of snoring origin and combined third classifier that fuses first classifier and second classification with mouth movement. Kopelman teaches systems associated with OSA. Senny teaches the second classifier (mouth position). MUELLER teaches anterior-posterior constriction, lateral constriction, and circular constriction. One of ordinary skills would have been motivated to combine Nwe , Kopelman, Senny and MUELLER that can provide improved precision to the distance to airway walls measurement (MUELLER [0183]). Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to TIRUMALE KRISHNASWAMY RAMESH whose telephone number is (571)272-4605. The examiner can normally be reached by phone. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Li B Zhen can be reached on phone (571-272-3768). The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /TIRUMALE K RAMESH/Examiner, Art Unit 2121 /Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Read full office action

Prosecution Timeline

Show 6 earlier events
Jun 27, 2025
Final Rejection mailed — §103
Jul 24, 2025
Examiner Interview Summary
Aug 22, 2025
Response after Non-Final Action
Dec 26, 2025
Request for Continued Examination
Jan 07, 2026
Response after Non-Final Action
Mar 25, 2026
Non-Final Rejection mailed — §103
Jun 24, 2026
Response Filed
Oct 07, 2026
Non-Final Rejection mailed — §103 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12792906
LEARNING APPARATUS, SIGNAL ESTIMATION APPARATUS, LEARNING METHOD, SIGNAL ESTIMATION METHOD, AND PROGRAM TO DEQUANTIZE
4y 3m to grant Granted Nov 10, 2026
Patent 12773776
DEEP REINFORCEMENT LEARNING FOR ROBOTIC MANIPULATION
5y 11m to grant Granted Oct 27, 2026
Patent 12769979
ARTIFICIAL INTELLIGENCE PLANNING METHOD AND ARTIFICIAL INTELLIGENCE PLANNING DEVICE
4y 11m to grant Granted Oct 20, 2026
Patent 12755142
CHIPS SUPPORTING CONSTANT TIME PROGRAM CONTROL OF NESTED LOOPS
5y 11m to grant Granted Oct 06, 2026
Patent 12699873
NEURAL NETWORK PROCESSING USING MIXED-PRECISION DATA REPRESENTATION
6y 6m to grant Granted Aug 04, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

4-5
Expected OA Rounds
28%
Grant Probability
52%
With Interview (+24.6%)
High
PTA Risk
Based on 51 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month