DETAILED ACTION
Applicant’s arguments, filed 02/13/2026, have been fully considered. The following rejections and/or objections are either reiterated or newly applied. They constitute the complete set presently being applied to the instant application.
Claims 1-20 are the current claims hereby under examination.
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 .
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-3 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar et. al. (“Estimating Respiratory Rate From Breath Audio Obtained Through Wearable Microphones” – cited by Applicant), hereinafter Kumar, and Musgrove (US 20200357518), further evidenced by Kogure (US 20200029832).
Regarding claim 1, Kumar discloses a method comprising:
obtaining at least one breathing audio sample of a user captured using earbuds worn by the user (Section 2, paragraphs 1-2, “… obtain breath samples of varying intensities … All data was recorded using microphone-enabled, near-range headphones, specifically Apple’s AirPods”);
converting the at least one breathing audio sample to a breathing spectrogram configured as an image (Figs. 2a-b);
processing the breathing data using a trained multi-task convolutional neural network (CNN) to identify image features of the image (Section 4b, paragraph 2, “The model, depicted in Figure 5, was trained with multiple and objective functions as a multi-task learning (MTL) network, where the tasks were RR estimation, heavy breathing detection, and noise detection”; Fig. 5 “feature extraction” image, wherein it is known by one of ordinary skill in the art that convolutional neural networks process and extract data from images);
predicting, using a deep neural network (DNN) layer of the multi-task CNN, a breathing rate and a breathing depth of the user using the image features (Fig. 5, outputs of respiration rate and heavy breathing,
outputting the breathing rate and the breathing depth of the user (Section 4b, paragraph 2, “The model, depicted in Figure 5, was trained with multiple objective functions as a multi-task learning (MTL) network, where the tasks were RR estimation, heavy breathing detection, and noise detection, represented by the three outputs mentioned above”).
Kumar discloses multiple layers of the models tested (Page 4, paragraphs 5-6 and 8-9) and Kogure shows that a neural network with an intermediate layer of a plurality of layers is a deep neural network and is trained by deep learning (Paragraph 0129). Kumar fails to disclose training a layer of the neural network separately.
However, Musgrove teaches a technique for preparing data for use in AI-based arrhythmia detection, wherein a neural network trains the intermediate layers separately then the subsequent layers of the neural network which accelerates training (Paragraph 0137). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Kumar and Kogure to train a layer of the neural network separately as taught by Musgrove to accelerate training of the neural network.
Regarding claim 2, Kumar as modified further discloses wherein the multi-task CNN is trained using multi-task learning in which the multi-task CNN is trained on multiple objective tasks in parallel, the multiple objective tasks comprising
a regression task associated with respiration rate (Section 4, paragraph 1, “The learning network, an end-to-end model, is not standard in that it simultaneously encompasses both regression and classification tasks.”; Section 4b, paragraph 3, “The individual losses from each task are given below, where concordance correlation coefficient (CCC) loss is used on the RR and RC outputs”) and
a classification task associated with respiration depth (Section 4b, paragraph 3, “… and weighted cross-entropy (CE) loss is used on the breath and noise classification tasks.).
Regarding claim 3, Kumar as modified further discloses wherein the multi-task learning uses a hybrid loss function that combines a regression loss for the regression task and a classification loss for the classification task (Section 4b, paragraph 4, “Additionally, a focal loss term was used for the breath detection task, and a convex mixture of all the losses after dynamic weight averaging, with weighting factor λ, was used as the MTL loss to train the network shown in Figure 5”).
Claims 4-6 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar, Musgrove, and Kogure as applied to claim 1 above, and further in view of Laredo et. al. (“Automatic model selection for fully connected neural networks”), hereinafter Laredo.
Regarding claims 4-5, Kumar discloses the multi-task CNN associated with a workout task as above. While Kumar discusses training different convolutional neural network models (Section 4B, paragraphs 1 and 7) based on a workout training dataset, Kumar fails to disclose choosing a neural network from a neural network pool and wherein an architecture of the CNNs is determined using a CNN neural architecture grid search.
However, Laredo teaches an algorithm for choosing one neural network from a selection of many neural networks in order to find the optimal neural network for a given dataset (Section 1, paragraph 6), wherein an architecture of each of the multiple CNNs in the CNN pool is determined using a CNN neural architecture grid search during training of the multiple CNNs (Section 2, paragraph 1; Section 4, paragraphs 2-6). Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Kumar and Musgrove to incorporate the CNN pool and architecture grid search of Laredo in order to find the optimal neural network for a given dataset.
Regarding claim 6, Kumar as modified further discloses determining that the participants are performing a workout during data collection (Section 2, paragraphs 3-4; Section 3, paragraph 1) and processes the data using a neural network trained on similar workout training data (Section 2, paragraphs 5-6), and Laredo discloses selecting a neural network based on a given dataset (Section 4, paragraph 2).
Claim 7 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar, Kogure, Musgrove, and Laredo as applied to claim 6 above, and further in view of Wisbey (US 20160051185).
Regarding claim 7, Kumar as modified fails to disclose determining the activity of a user through motion data.
However, Wisbey teaches a system for monitoring user activity through an earphone, wherein the earphone tracks user activity through a motion sensor (Paragraph 0089). One of ordinary skill in the art would have been capable of applying this known method of identifying a user activity via motion data collected by an earpiece in Wisbey to the earbuds of Kumar and the results of tracking user activity would have been predictable to one of ordinary skill in the art. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Kumar, Musgrove, and Laredo to incorporate the motion sensor of Wisbey, and the results of tracking user activity would have been predictable to one of ordinary skill in the art.
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Kumar, Kogure, and Musgrove as applied to claim 1 above, and further in view of Kariyawasan (WO 2020141999 – cited by Applicant).
Regarding claim 8, while Kumar discusses mel-filterbank energies of the breathing data, Kumar fails to explicitly disclose wherein the spectrogram is a mel-spectrogram.
However, Kariyawasan teaches a system for monitoring breathing of a patient from an earpiece (Fig. 1, earpiece breath monitoring device 110), wherein a sound classification model generates a mel-spectrogram from the input breath sound signal (Page 12, lines 1-4 ; Fig. 7) and a neural network is trained on the mel-spectrogram (Page 12, lines 6-15). Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method of Kumar and Musgrove with using a mel-spectrogram of Kariyawasan because it is a substitution of one known spectrogram for a known mel-spectrogram to yield the predictable result of training a neural network and determine breathing data parameters.
Claims 9-11 and 16-18 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar, Kogure, Kariyawasan, and Musgrove.
Regarding claims 9 and 16, Kumar discloses:
obtain at least one breathing audio sample of a user captured using earbuds worn by the user (Section 2, paragraphs 1-2, “… obtain breath samples of varying intensities … All data was recorded using microphone-enabled, near-range headphones, specifically Apple’s AirPods”);
convert the at least one breathing audio sample to a breathing spectrogram configured as an image (Figs. 2a-b);
process the breathing data using a trained multi-task convolutional neural network (CNN) to identify image features of the image (Section 4b, paragraph 2, “The model, depicted in Figure 5, was trained with multiple and objective functions as a multi-task learning (MTL) network, where the tasks were RR estimation, heavy breathing detection, and noise detection”; Fig. 5 “feature extraction” image, wherein it is known by one of ordinary skill in the art that convolutional neural networks process and extract data from images);
predicting, using a deep neural network (DNN) layer of the multi-task CNN, a breathing rate and a breathing depth of the user using the image features (Fig. 5, outputs of respiration rate and heavy breathing,
output the breathing rate and the breathing depth of the user (Section 4b, paragraph 2, “The model, depicted in Figure 5, was trained with multiple objective functions as a multi-task learning (MTL) network, where the tasks were RR estimation, heavy breathing detection, and noise detection, represented by the three outputs mentioned above”).
Kumar discloses multiple layers of the models tested (Page 4, paragraphs 5-6 and 8-9) and Kogure shows that a neural network with an intermediate layer of a plurality of layers is a deep neural network and is trained by deep learning (Paragraph 0129). While Kumar discusses processing the data (Section 4), Kumar fails to explicitly disclose any processing/memory structure configured to perform the method. Kumar also fails to disclose training a layer of the neural network separately.
However, Kariyawasan teaches a processor and storage device with computer executable instructions, which Kariyawasan discusses is useful to carry out the method of monitoring breathing signals (Page 3, lines 20-26). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the device of Kumar to incorporate the processor and storage device of Kariyawasan to carry out the method.
Musgrove teaches a technique for preparing data for use in AI-based arrhythmia detection, wherein a neural network trains the intermediate layers separately then the subsequent layers of the neural network which accelerates training (Paragraph 0137). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Kumar and Kariyawasan to train a layer of the neural network separately as taught by Musgrove to accelerate training of the neural network.
Regarding claims 10 and 17, Kumar as modified further discloses wherein the multi-task CNN is trained using multi-task learning in which the multi-task CNN is trained on multiple objective tasks in parallel, the multiple objective tasks comprising
a regression task associated with respiration rate (Section 4, paragraph 1, “The learning network, an end-to-end model, is not standard in that it simultaneously encompasses both regression and classification tasks.”; Section 4b, paragraph 3, “The individual losses from each task are given below, where concordance correlation coefficient (CCC) loss is used on the RR and RC outputs”) and
a classification task associated with respiration depth (Section 4b, paragraph 3, “… and weighted cross-entropy (CE) loss is used on the breath and noise classification tasks.).
Regarding claims 11 and 18, Kumar as modified further discloses wherein the multi-task learning uses a hybrid loss function that combines a regression loss for the regression task and a classification loss for the classification task (Section 4b, paragraph 4, “Additionally, a focal loss term was used for the breath detection task, and a convex mixture of all the losses after dynamic weight averaging, with weighting factor λ, was used as the MTL loss to train the network shown in Figure 5”).
Claims 12-14 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar, Kariyawasan, and Musgrove as applied to claims 9 and 16 above, and further in view of Laredo.
Regarding claims 12-14 and 19-20, Kumar as modified discloses the multi-task CNN associated with a workout task as above. While Kumar discusses training different convolutional neural network models (Section 4B, paragraphs 1 and 7) based on a workout training dataset, Kumar fails to disclose choosing a neural network from a neural network pool and wherein an architecture of the CNNs is determined using a CNN neural architecture grid search.
However, Laredo teaches an algorithm for choosing one neural network from a selection of many neural networks in order to find the optimal neural network for a given dataset (Section 1, paragraph 6), wherein an architecture of each of the multiple CNNs in the CNN pool is determined using a CNN neural architecture grid search during training of the multiple CNNs (Section 2, paragraph 1; Section 4, paragraphs 2-6). Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Kumar, Kariyawasan, and Musgrove to incorporate the CNN pool and architecture grid search of Laredo in order to find the optimal neural network for a given dataset.
Regarding claims 13 and 20, Laredo further teaches wherein an architecture of each of the multiple CNNs in the CNN pool is determined using a CNN neural architecture grid search during training of the multiple CNNs (Section 2, paragraph 1; Section 4, paragraphs 2-6).
Regarding claim 14, Kumar discloses determining that the participants are performing a workout (Section 2, paragraphs 3-4; Section 3, paragraph 1) and processes the data using a neural network trained on similar workout training data (Section 2, paragraphs 5-6), and Laredo discloses selecting a neural network based on a given dataset (Section 4, paragraph 2).
Claim 15 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar, Kariyawasan, Musgrove, and Laredo as applied to claim 14 above, and further in view of Wisbey.
Regarding claim 15, Kumar as modified fails to disclose determining the activity of a user through motion data.
However, Wisbey teaches a system for monitoring user activity through an earphone, wherein the earphone tracks user activity through a motion sensor (Paragraph 0089). One of ordinary skill in the art would have been capable of applying this known method of identifying a user activity via motion data collected by an earpiece in Wisbey to the earbuds of Kumar and the results of tracking user activity would have been predictable to one of ordinary skill in the art. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Kumar, Musgrove, and Laredo to incorporate the motion sensor of Wisbey, and the results of tracking user activity would have been predictable to one of ordinary skill in the art.
Response to Arguments
Applicant’s arguments, see page 8, filed 02/13/2026, with respect to the rejection(s) of claim(s) 1-3 under 35 U.S.C. §102(a)(1) have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Musgrove as described above.
Applicant asserts that Kumar fails to teach a CNN for identifying features of an image and using a deep neural network layer to predict a breathing rate and a breathing depth.
Kumar teaches using a time convolution LSTM network, a type of CNN, to estimate respiratory rate and heavy breathing and to detect noise (Figure 5 and Page 4, paragraph 1). The time-convolutional LSTM extracts features from an audio input (Figure 5, the image “feature extraction”). While Kumar does not explicitly state identifying image features, it is known within the art that convolutional neural networks process image data. As further evidenced by Kogure, the layers of Kumar are “deep neural network” layers.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Ahn et. al. (“Deep Elastic Networks with Model Selection for Multi-Task Learning”) teaches a method of neural network model selection based on the input.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NOAH MICHAEL HEALY whose telephone number is (703)756-5534. The examiner can normally be reached Monday - Friday 8:30am - 5:30pm ET.
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, Jason Sims can be reached at (571)272-7540. 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.
/NOAH M HEALY/Examiner, Art Unit 3791
/JASON M SIMS/Supervisory Patent Examiner, Art Unit 3791