DETAILED ACTION
Introduction
Applicant's submission filed on 02/20/2026 has been entered. Claims 1, 3-11, 13-15 and 17-25 are pending in the application and have been examined.
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 Amendment
The response filed on 02/20/2026 has been correspondingly accepted and considered in this Office Action. Claims 1, 3-11, 13-15 and 17-25 have been examined. New claims 24 and 25 have been added and examined. The Examiner reached out to the Applicant representative about the newly added independent claims 24 and 25 as being allowable as indicated in this Office Action, no agreement was reached.
Response to Arguments
Applicant's arguments filed 02/20/2026 have been fully considered as follows:
Applicant’s arguments with respect to claim 1 state that
“Salekin performs a binary determination of whether a target event is present in the clip as a whole, which serves as a "stop/go" gate rather than a label-based filter that extracts a discrete subset of waveforms for further analysis [Salekin, 0052].
Boyadjiev describes a voice authentication process that extracts acoustic features from a "voice sample" and applies a classifier to determine a probability of spoofing for that sample [Boyadjiev, 0045]. Boyadjiev does not perform its process on a subset of created by filtering using a first trained NN classifier.”
The examiner respectfully disagrees, Salekin teaches “Particularly, the processor 22 of the surveillance computer 20 is configured to execute program instructions corresponding to the DCNN audio tagging model 32 of the audio surveillance program 30 to determine whether the target audio event is present in the audio clip based on the plurality of HLD audio features HLD1, . . . , HLDN. As discussed above, in at least one embodiment, the processor 22 is configured to determine a classification output(s) C tag indicating whether the target audio event is present in the audio clip using a DCNN ” in Salekin, [0075], based on the audio features which are used to label each segment and once the audio event is detected in the audio clip, the audio clip is identified or filtered which is a subset of the plurality of audio clips to determine the position in time of the audio event. Therefore, Salekin teaches filtering the labelled waveforms of the predetermined length according to their label to obtain a plurality of audio events which is a subset of the plurality of waveforms each of the predetermined length. Boyadjiev processes the input waveforms which have passed the filtering by Salekin as taught in Boyadjiev, Fig. 1 and therefore, teaches determining a deepfake probability for each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length and therefore, the rejections of Claims 1 and 15 are rejected under 35 U.S.C. 103 are sustained and further updated accordingly.
In response to the art rejection(s) of the remainder of dependent claims are rejected under 35 U.S.C 103, in case said claims are correspondingly discussed and/or argued for at least the same rationale presented in Remarks filed 02/20/2026, Examiner respectfully notes as follows. For completeness, should the mentioned claims be likewise traversed for similar reasons to independent claims 1 and 15 correspondingly, Examiner respectfully directs Applicant to the same previous supra reasons provided in the response directed towards claims 1 and 15 correspondingly discussed above. For at least the same supra provided reasons, Examiner likewise respectfully disagrees, and Applicant's arguments have been fully considered but they are not persuasive.
Claim Rejections - 35 USC § 103
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1, 3-7, 12-15 are rejected under 35 U.S.C. 103 as being unpatentable over Salekin et. al, US PgPub. 2021/0005067 in view of Boyadjiev et. al, US PgPub. 2020/0035247 (cited in IDS) further in view of Gopala et. al., US PgPub. 2021/0074305.
Regarding claim 1, Salekin teaches a method comprising: dividing all or a part of an input waveform into a plurality of waveforms each of a predetermined length (see Salekin, [0073] the processor 22 is configured to receive audio surveillance signals in the form of audio clips having a predetermined length (e.g., 30 seconds). In some embodiments, the processor 22 is configured to receive the audio surveillance signals in the form of an audio stream and divide the audio stream into audio clips having a predetermined length (e.g., 30 seconds)); labelling each of the plurality of waveforms each of a predetermined length by a first trained NN classifier (see Salekin, [0074] determines the audio features for each of the audio clips (label)); filtering the labelled waveforms of the predetermined length according to their label to obtain a plurality of audio events which is a subset of the plurality of waveforms each of the predetermined length (see Salekin, [0075] determine using the DCNN audio tagging model(based on the labelled waveforms), if the target event in the present in the audio clip (filtering)). However, Salekin fails to teach determining a deepfake probability for each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length with a second trained NN classifier, wherein the deepfake probability indicates a probability that the input waveform has been altered by artificial intelligence techniques and/or distorted by artificial intelligence techniques or has been completely generated by artificial intelligence techniques; combining the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length.
However, Boyadjiev teaches determining a deepfake probability for each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length with a second trained NN classifier, wherein the deepfake probability indicates a probability that the input waveform has been altered by artificial intelligence techniques and/or distorted by artificial intelligence techniques or has been completely generated by artificial intelligence techniques (see Boyadjiev, Fig. 1 and [0015] The system may apply each multi-dimensional acoustic feature vector in the plurality of multi-dimensional acoustic feature vectors to a corresponding trained multi-dimensional acoustic feature vector machine learning classifier to determine a probability of spoofing of each multi-dimensional acoustic feature vector); combining the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length(see Boyadjiev, [0015] The system may then determine an overall probability of spoofing for the voice sample based on the probabilities of spoofing of the plurality of multi-dimensional acoustic feature vectors, Boyadjiev, [0068] At 705, the multi-dimensional acoustic feature vector CNN 121A, 121B, 121C or 121D may determine an overall probability of spoofing for the multi-dimensional acoustic feature vector, based on the probability of spoofing for each multi-dimensional acoustic feature vector).
Salekin and Boyadjiev are considered to be analogous to the claimed invention because they relate to identification of audio content. 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 teachings of Salekin on determining a target audio even with the determination of if an authentication request is a spoofed voice such as a synthesized voice or a converted voice teachings of Boyadjiev to improve voice authentication(see Boyadjiev,[0001-0002]).
However, Salekin in view of Boyadjiev fail to teach generating a modified audio waveform that is a modification of the input audio waveform based upon the combination of the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length .
However, Gopala teaches generating a modified audio waveform that is a modification of the input audio waveform based upon the combination of the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length (see Gopala, [0111] the output 414 is a SoftMax output ( combination of deepfake probabilities), such as a value that indicates fake or real, or a probability. Gopala [0100] the classifier output 216 is processed and provided with additional information regarding real or fake along with the audio content ( modification of audio waveform)).
Salekin, Boyadjiev and Gopala are considered to be analogous to the claimed invention because they relate to identification of fake audio content. 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 teachings of Salekin in view of Boyadjiev on determining if the authentication request is a spoofed voice such as a synthesized voice or a converted voice with the identification teachings of fake audio content teachings of Gopala to reduce dispersion of fake audio content in the online platforms(see Gopala,[0003-0004]).
Regarding claim 3, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Boyadjiev further teaches wherein the audio waveform relates to media content (see Boyadjiev, [0023] The voice samples 140 may be provided in audio files and generated from one or more frames of digital audio. The voice samples 140 may be provided by any number of sources ). The same motivation to combine as claim 1 applies here.
Regarding claim 4, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Salekin further teaches determining a candidate spectrogram for each of the plurality of waveforms of the predetermined length, wherein each of the candidate spectrograms is labelled by the first trained NN classifier (see Salekin, [0074] determines the audio features for each of the audio clips (label)), the labelled spectrograms are filtered according to their label to obtain a plurality of audio event spectrograms which is a subset of the candidate spectrograms (see Salekin, [0075] determine using the DCNN audio tagging model, if the target event in the present in the audio clip (filtering)). Boyadjiev teaches the deepfake probability is determined for each of the plurality of audio event spectrograms of the subset of the candidate spectrograms with the second trained NN classifier (see Boyadjiev, Fig. 1 and [0015] The system may apply each multi-dimensional acoustic feature vector in the plurality of multi-dimensional acoustic feature vectors to a corresponding trained multi-dimensional acoustic feature vector machine learning classifier to determine a probability of spoofing of each multi-dimensional acoustic feature vector). Gopala teaches the deepfake probabilities of each of the plurality of audio event spectrograms of the subset of the candidate spectrograms are combined (see Gopala, [0111] the output 414 is a SoftMax output ( combination of deepfake probabilities), such as a value that indicates fake or real, or a probability), and the modified waveform is generated based upon the combination of the deepfake probabilities of each of the plurality of audio event spectrograms of the subset of the candidate spectrograms (Gopala [0100] the classifier output 216 is processed and provided with additional information regarding real or fake along with the audio content ( modification of audio waveform)). The same motivations to combine as claim 1 applies here.
Regarding claim 5, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Salekin further teaches wherein the first trained NN classifier is a trained DNN classifier (see Salekin, [0027, 0049] the BLSTM classifier model 38 (first trained NN) is configured to identify the boundaries and/or positions in time of the detected target audio event in the audio clip and the training of the DCNN).
Regarding claim 6, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Boyadjiev further teaches performing audio source separation on the input audio waveform to obtain a vocal waveform or a speech waveform and taking the vocal waveform or the speech waveform as the input waveform (see Boyadjiev, [0015] To determine the probability of voice spoofing, the system may extract an acoustic feature from a voice sample of a user. For example, an acoustic feature may be a part of the voice sample that contains a human voice separated from background noise or pauses. The system may convert the acoustic feature into a plurality of multi-dimensional acoustic feature vectors. In an example, a multi-dimensional acoustic feature vector may represent voice attributes in the voice sample in a visual form(spectrogram)). The same motivations to combine as claim 1 applies here.
Regarding claim 7, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Boyadjiev further teaches wherein the second trained NN classifier is a trained DNN classifier (see Boyadjiev, [0016-0017] discusses the trained classifier comparing the particular features(labels) of the multi-dimensional acoustic feature vectors(spectrogram) relevant for authentication to output an accurate classification (by trained classifier) of the multi-dimensional acoustic feature vector for detecting the probability of spoofing. [0027] The plurality of multi-dimensional acoustic feature vector CNNs may produce extracted attributes 160 from the plurality of multi-dimensional acoustic feature vectors 151. Examples of extracted attributes 160 may include a probability of spoofing of a multi-dimensional acoustic feature vector(spectrograms), the differences between the multi-dimensional acoustic feature vectors of a spoofed sample and a human voice and the like. Fig. 4). The same motivations to combine as claim 1 applies here.
Regarding claim 13, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Gopala further teaches wherein generating the modified audio waveform includes overlaying a warning message over the input audio waveform based on the combination of deepfake probability probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length (see Gopala, [0111] the output 414 is a SoftMax output ( combination of deepfake probabilities), such as a value that indicates fake or real, or a probability. see Gopala, [0100] Next, the computer system may classify, based at least in part on an output of the predetermined neural network, the audio content (operation 216) as being fake or real, where the fake audio content is, at least in part, computer-generated. Furthermore, the computer system may selectively perform a remedial action (operation 218) based at least in part on the classification. For example, the remedial action may include one or more of: providing a warning associated with the audio content(overlay warning message over the audio); providing a recommendation associated with the audio content; or filtering at least a portion of the audio content (such as removing or changing at least the portion of the audio content).[0119] For example, the reality defender, which may be based at least in part on deep learning, may be implemented as a Web browser plugin or a software application that can notify users of suspected deception (such as fake media) in real time. Notably, the reality defender may analyze suspected media and may provide warnings or alerts, and/or may filter out identified fake media.). the same motivation to combine as claim 1 applies here.
Regarding claim 14, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. Gopala further teaches outputting a warning based on the combination of deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length (see Gopala, [0111] the output 414 is a SoftMax output ( combination of deepfake probabilities), such as a value that indicates fake or real, or a probability. see Gopala, [0100] Next, the computer system may classify, based at least in part on an output of the predetermined neural network, the audio content (operation 216) as being fake or real, where the fake audio content is, at least in part, computer-generated. Furthermore, the computer system may selectively perform a remedial action (operation 218) based at least in part on the classification. For example, the remedial action may include one or more of: providing a warning associated with the audio content(overlay warning message over the audio); providing a recommendation associated with the audio content). The same motivation to combine as claim 1 applies here.
Regarding claim 15, is directed to an electronic device claim corresponding to the method claim presented in claim 1 and is rejected under the same grounds stated above regarding claim 1.
Claims 8-9, 17-18 and 22-23 are rejected under 35 U.S.C. 103 as being unpatentable over Salekin et. al, US PgPub. 2021/0005067 in view of Boyadjiev et. al, US PgPub. 2020/0035247 (cited in IDS) further in view of Gopala et. al., US PgPub. 2021/0074305 further in view of Kong, Q, et al. "Audio set classification with attention model: A probabilistic perspective." 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2018.
Regarding claim 8, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. However, Salekin in view of Boyadjiev further in view of Gopala fails to teach wherein determining the deepfake probability for each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length includes determining an intrinsic dimension probability value of each of the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length.
However, Kong teaches determining an intrinsic dimension probability value of each of the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length (see Kong, Fig. 1 describes classification result for each instance and the probability measure for each instance; Kong, sect. 2 In Audio Set classification, a bag is a collection of L features from an audio clip. Each instance xnl ∈ RM is a feature, where M is the dimension of the feature. The label of a bag is dn ∈ {0, 1} K where K is the number of audio classes and 0 and 1 represent the negative and positive label, respectively. Kong, sect 3.4 equation 6 ( intrinsic probability)).
Salekin, Boyadjiev and Gopala teaches determining probability of spoofing. Ivry teaches methods to tagging and classification of audio set using probabilistic methods(see Kong, sect. 1). Using the known technique of audio classification using attention model method as taught by Kong, to determine probabilistic value of the audio event in the reference Salekin, Boyadjiev and Gopala, such as improved deepfake detection would have been obvious to one of ordinary skill in the art.
Regarding claim 9, Salekin in view of Boyadjiev further in view of Gopala further in view of Kong teaches the method of claim 8. Kong further teaches wherein a respective intrinsic dimension probability value is determined based on: a ratio of an intrinsic dimension of (see Kong, sect 3.4 Then for each bag Bn and x ∈ Bn, we may define the probability measure of any instance x of the k-th class as: equation 6, which is the ratio of µ({x}) and µ(Bn) where µ({x}) and µ(Bn) are the measure of {x} and Bn, respectively).
Regarding claim 17, Salekin in view of Boyadjiev further in view of Gopala further in view of Kong teaches the method of claim 8. Kong further teaches wherein the deepfake probability for each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length is an average of a fake probability value determined as a result of using the second trained NN classifier for a respective one of the audio events (see Kong, sect. 2.1 , the bag level classifier F is obtained by using the sum as the aggregation rule (equation 1) (Average pooling) ) and the intrinsic dimension probability value of the respective one of the audio events(see Kong, sect 3.4, equation (6), Kong, Fig. 2 ).
Regarding claim 18, Salekin in view of Boyadjiev further in view of Gopala further in view of Kong teaches the method of claim 8. Kong further teaches wherein the deepfake probability for each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length is a maximum of a fake probability determined as a result of using the second trained NN classifier for a respective one of the audio events (see Kong, sect. 2.2 , The maximum selection [11] states that the prediction of a bag is the maximum classification value of each instance in the bag described (equation 2) (Max pooling) ))and the intrinsic dimension probability value of the a respective one of the audio events(see Kong, sect 3.4, equation (6), Kong, Fig. 2 ).
Regarding claim 22, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. However, Salekin in view of Boyadjiev further in view of Gopala fails to teach wherein labeling each of the plurality of waveforms of the predetermined length includes assigning labels from a predefined set of audio categories, and filtering the labelled waveforms of the predetermined length includes selecting waveforms labeled as human speech.
However, Kong teaches wherein labeling each of the plurality of waveforms of the predetermined length includes assigning labels from a predefined set of audio categories, and filtering the labelled waveforms of the predetermined length includes selecting waveforms labeled as human speech (see Kong, sect 2.3 indicates audio tagging and audio event detection(human speech) based on predefined audio categories as described [6] and [14]). The same motivation to combine as claim 8 applies here.
Regarding claim 23, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 1. However, Salekin in view of Boyadjiev further in view of Gopala fails to teach wherein combining the deepfake probabilities includes calculating a weighted average of the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length, and the weights are assigned based on whether a respective one of the audio events contains human speech.
However, Kong teaches wherein combining the deepfake probabilities includes calculating a weighted average of the deepfake probabilities of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length, and the weights are assigned based on whether a respective one of the audio events contains human speech (see Kong, sect 2.3 The weighted collective assumption asserts that each instance contributes independently but not necessarily equally to the label of a tag. This is achieved by incorporating a weight function w(x) into the collective assumption and indicates audio tagging and audio event detection(human speech) based on predefined audio categories as described [6] and [14]). The same motivation to combine as claim 8 applies here.
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Salekin et. al, US PgPub. 2021/0005067 in view of Boyadjiev et. al, US PgPub. 2020/0035247 (cited in IDS) further in view of Gopala et. al., US PgPub. 2021/0074305 further in view of Stamatopoulos et. al., US PgPub. 2019/0088367.
Regarding claim 10, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 4. However, Salekin in view of Boyadjiev further in view of Gopala fails to teach each of the plurality of audio event spectrograms of the subset of the candidate spectrograms includes determining a correlation probability value of each of the plurality of audio event spectrograms of the subset of the candidate spectrograms.
However, Stamatopoulos teaches each of the plurality of audio event spectrograms of the subset of the candidate spectrograms includes determining a correlation probability value of each of the plurality of audio event spectrograms of the subset of the candidate spectrograms (see Stamatopoulos, [0415] discusses At step 2753, a cross correlation function(correlation probability) is determined between every frame and the normalized filtered response( each audio event spectrogram). FIG. 33 illustrates the cross correlation function( each audio event spectrogram) determined using the frame and the normalized filtered response ).
Salekin, Boyadjiev and Gopala teaches determining probability of spoofing. Stamatopoulos teaches methods for detecting crackles using cross correlation functions (see Stamatopoulos, [0415]). Using the known technique of cross correlation method as taught by Stamatopoulos, to determine correlation probability value of the audio event in the references of Salekin, Boyadjiev and Gopala, such as improved deepfake detection would have been obvious to one of ordinary skill in the art.
Allowable Subject Matter
Claims 11 and 19-21 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Claim 11 is objected because the prior art of record fails to teach or suggest a normalized cross-correlation between a resized stored real audio event spectrogram of a recording noise floor and noise-only parts of a respective one of the audio event spectrograms, in combination with all of the elements of claim 10 from which it is dependent. The closet prior art of record, Stamatopoulos teaches the method of claim 10 to determine a correlation probability function. Additionally, Salekin in view of Boyadjiev further in view of Gopala teach the method of claim 4. However, Salekin in view of Boyadjiev further in view of Gopala further in view of Stamatopoulos fails to reasonably teach or suggest “a normalized cross-correlation between a resized stored real audio event spectrogram of a recording noise floor and noise-only parts of a respective one of the audio event spectrograms” as currently claimed.
Claim 19 is objected because the prior art of record fails to teach or suggest includes determining a correlation probability value of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length in combination with all of the elements of claim 8 from which it is dependent. The closet prior art of record, Kong teaches the method of claim 8 to determine an intrinsic dimension probability value. Additionally, the closet prior art of record, Stamatopoulos teaches the method of claim 10 to determine a correlation probability function. Additionally, Salekin in view of Boyadjiev further in view of Gopala further in view of Kong teach the method of claim 8. However, Salekin in view of Boyadjiev further in view of Gopala further in view of Kong fails to reasonably teach or suggest “includes determining a correlation probability value of each of the plurality of audio events of the subset of the plurality of waveforms each of the predetermined length” as currently claimed.
Claim 20 is objected to due to its dependence on claim 19.
Claim 21 is objected to due to its dependence on claim 19.
Claims 24 and 25 are allowed for the reasons stated below.Claim 24 includes claim 1 with the allowable subject matter from claim 11 and are allowable for the similar reasons of claim 11.
Claim 25 includes claim 1 with the allowable subject matter from claim 19 and are allowable for the similar reasons of claim 19.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Shiloh et. al, US Patent 11,948,599 teaches detecting audio events based on the labels and filtering based on the classification scores ( see Shiloh, Fig. 2).
Traynor et. al. US PgPub 2022/0036904 teaches the cross-sectional area estimates for one of the bigrams from the ideal feature set and cross-sectional area estimates were then converted to their approximate diameters may be used to evaluate the correlation between two sounds. The divergence between the cross-sectional area estimates and/or diameters may be used to distinguish between different speakers of a sound, or establish if an audio sample was organically generated or digitally constructed (see Traynor, [0075]).
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 NANDINI SUBRAMANI whose telephone number is (571)272-3916. The examiner can normally be reached Monday - Friday 12:00pm - 5:00 pm EST.
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, Bhavesh M Mehta can be reached at (571)272-7453. 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.
/NANDINI SUBRAMANI/ Examiner, Art Unit 2656 /BHAVESH M MEHTA/Supervisory Patent Examiner, Art Unit 2656
Prosecution Insights