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 .
Information Disclosure Statement
Acknowledgment is made of the information disclosure statements filed on 4/10/2026. It is noted by the Examiner that all of the references were considered.
Response to Arguments
35 USC 101
Applicant's arguments filed 4/10/2026 are considered in this office action, but argument are not percussive and 101 rejection maintained.
Applicant Asserts:
“The Claims Could Not Practically Be Performed in the Human Mind
The Examiner characterizes the claimed method as a mental process performable by a
human drawing frequency representations on paper and analyzing them mentally. (Office Action at p. 5-6). Applicant respectfully submits that this characterization is not objectively reasonable. The claimed method recites receiving a digital representation of an audio signal, applying frequency domain transforms, determining mel-frequency cepstral coefficients, and computing pitch indicators from digital audio data. Even if one could theoretically imagine a human performing some of these steps in isolation, the practical reality is that these operations cannot be performed by the human mind. Courts have recognized that claims directed to specific technical processes involving digital data are not abstract merely because they could theoretically be characterized as mental steps. (See MEPEP 2106.04(a)(2)(11l)(A) (A Claim With Limitation(s) That Cannot Practically be Performed in the Human Mind Does Not Recite a Mental Process). Accordingly, Applicant respectfully submits that the claims are not directed to a mental process.” {Remark Page 6, paragraph 1}
Examiner Note: Human can divide audio data into frames, frames representing a time intervals which has a set of value called vector/feature vector, representing features of the utterance portion within a frame. Human can draw (x-y axis) time frequency representation audio signal/frame on a piece of paper, and determine mel-cepstral, pitch data in the graph, calculate FFT for same audio signal/frames. By recognizing pitch, tone, mel data human can determined voice activity present or not.
Applicant Asserts:
“The Cascaded Two-Stage Architecture Reflects a Technological Improvement Under MPEP 2106.04(d)(l)” {Remark page 6, last paragraph}
Examiner Note: The claim does not reflect any Cascaded architecture. Sequential audio processing cannot be cascaded architecture. It’s just a two steps traditional Voice Activity Detection (VAD) processing. First stage, determining user speech and then analyzing identified speech in time domain and in frequency domain. Then, FFT transform is needed to convert from time domain to frequency domain. Next analyze pitch, tone, mel-ceptral in the frames to determine voice activity.
Applicant further Asserts:
First, MPEP 2106.04(d)(l) expressly states that "the claim itself does not need to explicitly
recite an improvement described in the specification ( e.g., 'thereby increasing the bandwidth of the channel')." The Examiner's position that the claims must explicitly recite conservation of processor load, bandwidth, or memory directly contradicts this MPEP language. The Examiner has not acknowledged or addressed this specific provision. {Remark Page 7, 1st Para.}
Examiner Note:
(“MPEP 2106.04(d)(l) … Conversely, if the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner has determined there is nothing in the clam that improves technology. Second, if the specification sets forth an improvement in technology, the claim must be evaluated to ensure that the claim itself reflects the disclosed improvement. That is, the claim includes the components or steps of the invention that provide the improvement described in the specification.
Regarding rejection to claims under 35 U.S.C. §103, Applicant argument (Remarks, pages 8-10) have been considered but are not percussive, therefore 35 U.S.C. §103 rejection maintained.
Applicant Asserts: “This is a direct final determination per frame, not a candidate identification for further processing by a second stage. Applicant respectfully submits that the cited passage shows precisely the single-stage architecture that Applicant previously argued York discusses.”{Remark page 9, paragraph 1-3}.
“Claim 1 recites a cascaded two-stage architecture: a first stage applies a first frequency domain indicator to identify a candidate speech duration, and a second stage is then applied specifically to that identified candidate speech duration to assess whether it contains speech. Support for this architecture is found at paragraphs [0026]-[0027] of the Application as filed.” {Remarks, pages 9}
Examiner Response: Examiner respectfully disagree with applicant assertion. Both York and Otani overlap claim invention smoothly. York teaches voice activity, voice data frame length in a time domain and FFT transform/conversion, and York determines frequency domain features include mel-frequency cepstral coefficients. Otani also teaches Fourier transform algorithm converts a time series of samples into a set of components in the frequency domain, pitch structure and voice flatness comparison in frame level and the voice/noise discriminator 33f compares the flatness factor of each frame with a predetermined threshold. Therefore, York and Otani overlap first stage and second stage smoothly.
Claim does not reflect any cascaded two-stage architecture: Claim present sequential voice processing in two steps. There is no architecture recited in the claims. York teaches first stage, determining user speech and then analyzing identified speech in time domain and in frequency domain. “… speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. …”col. 46 and col. 47, lines 56-57 and lines 1-10) York teaches (“(218) … speech detector 852 divides an input data stream into frames (e.g., portions) of a predetermined length, e.g., 25, 50, 100, 200, 300, 400, or 500 milliseconds. In some examples, speech detector 852 determines whether each frame indicates that a user is speaking (indicates user speech). …” Col. 47, lines 1-13) York even teaches determining whether the first data stream indicates user speech includes analyzing the time domain and/or frequency domain features of the first data stream to determine whether such features indicate human speech and frequency domain features include mel-frequency cepstral coefficients, and comparing the data stream to a human speech model) may be employed to determine whether a data stream indicates user speech. For example, speech detector 852 processes the first data stream to determine that a 2 second portion of the data stream indicates user speech. For example, speech detector 852 determines that each frame of a consecutive series of frames (having a collective 2 second duration) indicates user speech.” in col. 47, lines 33- 54)
Examiner interprets second stage explicitly teaches by Otani. Please see FIG. 1-20, 1(A), (B), 14(B), FIG. 16 and FIG. 17 Otani teaches in paragraph (“[0038] …In the case of FFT, the Fourier transform algorithm converts a time series of samples into a set of components in the frequency domain, i.e., the frequency spectrum of the signal. Suppose now that a time-domain data stream x for one frame period is given. The given stream is converted to a frequency-domain dataset X=(X[k].vertline.k=1, 2, . . . N), where k is frequency and N is the total number of subdivided (i.e., discretized) frequency bands.”). Otani teaches pitch structure in paragraph [0055] … Pitch structures indicate the tone height, which comes from the frequency of vocal chord vibration. …”). Otani even teaches voice flatness comparison in frame level (“Talkspurts can be identified by calculating flatness factors of spectrums and comparing them (the voice/noise discriminator 13 actually compares the flatness factor with a predetermined threshold).”) Otani teaches (“[0114] (S14) The voice/noise discriminator 33f compares the flatness factor of each frame with a predetermined threshold. Through this comparison the voice/noise discriminator 33f determines whether the frame in question is speech or noise, and it sets an appropriate flag to indicate the result.”)
Applicant further Asserts: (“The Office Action's Citation to York's STT Processing Module to Teach the VAD Second Stage Is a Mischaracterization of York's Disclosed Architecture
The Office Action cites York at col. 35, lines 43-53 in support of the second stage limitation.
This passage describes York's speech-to-text processing module 730, which performs Fourier
transforms to extract spectral features for speech recognition. This module operates on speech that has already been identified and is not a voice activity detection component. Applicant respectfully submits that citing a speech recognition module to teach a VAD second stage is inconsistent with York's own disclosed architecture.”
Examiner Response: Even if the claim mapping somehow shows “Mischaracterization of York's Disclosed Architecture” (which Examiner does not concede), performs Fourier
transforms to extract spectral features for speech recognition shows spectral features speech/voice (emphasis added). Please, see examiner detail explanation above regarding first stage and second stage.
“Applicant further notes that claim 12 is rejected over York and Visser without Otani, which
appears inconsistent with the rejection of claim 1 upon which claim 12 depends. Applicant
respectfully requests clarification on this point.” (Remak page 10, last paragraph)
Examiner Response: Claim 12 header has typographical error, but motivation paragraph recites “modify the York, and Otani to incorporate the teachings of Visser …”. Examiner correct typo, now. “Claim 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over York et al. and Otani in view of Visser et al. US 20130282373 A1.”
Applicant further Asserts:
The Motivation to Combine Is Insufficient: The stated motivation to combine, namely that the combination enables a model to "set an appropriate flag to indicate the result," is insufficient. York already makes speech/no-speech determinations. The stated motivation does not explain why a skilled artisan would modify York's single-stage architecture to incorporate any portion of Otani, and it does not identify any problem the combination solves, and is identical boilerplate applied to every claim grouping in the Office Action regardless of the specific limitations at issue.”
Examiner Response: Examiner respectfully disagree with applicant assertion. York and Otani both prior arts are related to detecting voice signal, calculate FFT, analyzing pitch, tone, mel data of audio/frames. One could have been motivated to do so modify the York to incorporate the teachings of Otani in order to provide a voice activity detector that detects talkspurts/voice activity in a given signal at a high accuracy so as to improve the quality of voice communication.
Therefore, combination of these two arts are obvious.
Therefore, examiner maintained all previous rejection.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim(s) 1, 19, and 20, the limitation(s) of “receive audio data”, “applying”, and “applying”, as drafted, are processes that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. More specifically, the mental process of a human receiving audio data and drawing audio frequency representation on a piece of paper based on human knowledge, and evaluating frequency value would be speech duration or not, representing this as two or three steps would mental process of a human. Also, human can compare candidate speech duration values with predefined threshold/rule/FFT apply and evaluate/assess speech/noise is present or not.
If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the --Mental Processes-- grouping of abstract ideas. Accordingly, the claim(s) recite(s) an abstract idea.
This judicial exception is not integrated into a practical application because the recitation of “receiver circuit”, “processor circuit” and “memory circuit” in claim 20, reads to generalized computer components, based upon the claim interpretation wherein the structure is interpreted using [0011], [0031], [0064] in the specification. Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim(s) is/are directed to an abstract idea.
The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional element of using generalized computer components to receive, transform, determine, determine, determine, determine, and indicate amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The claim(s) is/are not patent eligible.
With respect to claim(s) 2, the claim(s) recite(s), “establishing respective frames defining specified durations within the digital representation of the audio signal” reads on a human mind creating/drawing frame by frame corresponding to audio signal No additional limitations are present.
With respect to claim(s) 3, the claim(s) recite(s), “receiving a digital representation of an audio signal comprises receiving a streamed representation” reads on a human mind obtain real time conversation or any continuous signal and this continuous audio signal can be represented to as a frame. No additional limitations are present.
With respect to claim(s) 4, the claim(s) recite(s) “first frequency-domain indicator includes determining a representation of a frequency dispersion of spectral components of the digital representation of the audio signal,” , reads on a human spectral analysis of frequency domain frames. No additional limitations are present.
With respect to claim(s) 5, the claim(s) recite(s),” comparing the determined representation of the dispersion with a first threshold“ which reads on a human looking at the two values to determine if the calculated value meets the threshold value. No additional limitations are present.
With respect to claim(s) 6, the claim(s) recite(s), “adjusting the first threshold“ which reads on a human adjusting or tunning threshold value based on spectral centroids. No additional limitations are present
With respect to claim(s) 7, the claim(s) recite(s), “adjusting the first threshold“ which reads on a human adjusting or tunning threshold value based on frame where no speech present. No additional limitations are present
With respect to claim(s) 8 and 9, the claim(s) recite(s), mathematical calculation of the cepstrum converts a signal to the frequency domain through Fourier transform, takes the logarithm, then applies another Fourier transform to find inverse function. No additional limitations are present
With respect to claim(s) 10, the claim(s) recite(s), “comparing the determined central tendency to a threshold” comparing two value and determine threshold is exceed or within limit. This is mental activity and comparing two numbers is mathematical algorithm at the same time. Therefore, this claim recites abstract idea. No additional limitations are present
With respect to claim(s) 11-13, the claim(s) recite(s), “MFC indicator” defines audio signal and how scattered are frequency component in the spectral graph, and further analyze these graphs based on speech present prediction. Human analyzing data is human activity which are abstract idea. Therefore, this claim recites abstract idea. No additional limitations are present
With respect to claim(s) 14, the claim(s) recite(s), “second stage comprises both an MFC indicator and pitch indicator” which reads on a human to subsequent steps/stages are there to analyze portion of audio has speech or not, and system analyzes the MFCC to identify the pitch of the words. No additional limitations are present.
With respect to claim(s) 15 and 16, the claim(s) recite(s) duration of speech sending to another system as it receiving partially, which reads on a human portion of speech send to external device as it receives. These claims are pre and post solution activity. No additional limitations are present
With respect to claim(s) 17 and 18, the claim(s) recite(s) “applying a third stage comprising at least one temporal indicator to assess”, which reads on a human predefined temporal relationship and comparing with candidate speech duration values or not. No additional limitations are present
These claims further do not remedy the judicial exception being integrated into a practical application and further fail to include additional elements that are sufficient to amount to significantly more than the judicial exception.
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.
Claim 1, 2, 3, 4, 11, 15, 16, 17, 18 19, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over York et al. US 11705130 B2 in view of Otani et al. US 20050108004 A1.
Regarding Claim 1, York teaches
1. A machine-implemented method for detecting voice activity, the method comprising: receiving a digital representation of an audio signal. York teaches (“(9) … obtain one or more data streams from one or more sensors; determine, based on the one or more data streams, whether a user associated with the electronic device is speaking; in accordance with a determination that the user is speaking: …” col. 2, lines 42-45) (“(218) In some examples, determination module 850 includes speech detector 852 configured to process data streams. In some examples, speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. …”col. 46, lines 56-57 ) by York et al. US 11705130 B2
applying a first stage comprising determining a first frequency-domain indicator from the digital representation of the audio signal to identify a candidate speech duration; and
York teaches (“(218) In some examples, determination module 850 includes speech detector 852 configured to process data streams. In some examples, speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. In some examples, speech detector 852 divides an input data stream into frames (e.g., portions) of a predetermined length, e.g., 25, 50, 100, 200, 300, 400, or 500 milliseconds. In some examples, speech detector 852 determines whether each frame indicates that a user is speaking (indicates user speech). For example, speech detector 852 determines a probability of user speech for each frame and/or makes a “yes” or “no” decision regarding the presence of user speech for each frame. In this manner, speech detector 852 can identify the boundaries of user speech in a data stream and determine the duration(s) during which a user is speaking or not speaking.” Col. 47, lines 1-13) by York et al. US 11705130 B2
the second stage comprising determining at least one of a mel- frequency cepstral (MFC) indicator or a pitch indicator from the digital representation of the audio signal corresponding to the identified candidate speech duration to assess whether the identified candidate speech duration contains speech. York teaches (“(169) STT processing module 730 includes one or more ASR systems 758. The one or more ASR systems 758 can process the speech input that is received through I/O processing module 728 to produce a recognition result. Each ASR system 758 includes a front-end speech pre-processor. The front-end speech pre-processor extracts representative features from the speech input. For example, the front-end speech pre-processor performs a Fourier transform on the speech input to extract spectral features that characterize the speech input as a sequence of representative multi-dimensional vectors. …” COL. 35, lines 43-53) (“(221) In some examples, determination module 850 determines whether a first data stream obtained from a speech sensor (or a portion thereof) indicates user speech. In some examples, determining whether the first data stream indicates user speech includes analyzing the time domain and/or frequency domain features of the first data stream to determine whether such features indicate human speech. Exemplary time domain features include zero crossing rates, short-time energy, spectral energy, spectral flatness, and autocorrelation. Exemplary frequency domain features include mel-frequency cepstral coefficients, linear predictive cepstral coefficients, and mel-frequency discrete wavelet coefficients. In addition to analyzing such features, one of skill in the art will appreciate that any other suitable technique (e.g., comparing the data stream to a human speech model) may be employed to determine whether a data stream indicates user speech. For example, speech detector 852 processes the first data stream to determine that a 2 second portion of the data stream indicates user speech. For example, speech detector 852 determines that each frame of a consecutive series of frames (having a collective 2 second duration) indicates user speech.” col. 47, lines 33- 54) (“(230) … in some examples, determination module 850 determines whether a length (e.g., number of words, length of audio) of a previous notification is below a threshold length. In some examples, if the length is determined to be below the threshold length, …”) (“(234) In FIG. 8A, the user is determined to be not speaking. Specifically, during a 7 second time window after the text message was received, a 2 second duration of no user speech is determined. During that duration, device 810 provides the audio output “Mom says ‘where are you?’””) (“(278) In some examples, it is determined, based on obtained data stream(s), whether a user associated with an electronic device (e.g., devices 800 and 810) is speaking (e.g., by determination module 850). Determining whether a user is speaking is performed according to any of the techniques discussed above with respect to FIGS. 8A-8C. For example, determining whether a user is speaking includes determining whether a data stream obtained from a vibration sensor (e.g., bone conduction microphone) indicates user speech for a predetermined duration (e.g., 0.2, 0.3, 0.4, or 0.5 seconds). …”) by York et al. US 11705130 B2
York does not explicitly teach corresponding to the identified candidate speech duration to assess whether the identified candidate speech duration contains speech in second stage.
Otani teaches:
applying a second stage pitch of the audio signal to the candidate speech duration identified using the first frequency- domain indicator of the first stage, corresponding to the identified candidate speech duration to assess whether the identified candidate speech duration contains speech. Otani teaches Fourier transform algorithm converts a time series of samples into a set of components in the frequency domain. (“[0149] (S42) The power spectrum calculator 64a applies FFT on the input sound signal and supplies the resulting power spectrum to the talkspurt detector 64b.”) (“[0038] In the spectral analysis approach, the power spectrum of a signal is calculated with fast Fourier transform (FFT), wavelet transform, or other known algorithms. In the case of FFT, the Fourier transform algorithm converts a time series of samples into a set of components in the frequency domain, i.e., the frequency spectrum of the signal. Suppose now that a time-domain data stream x for one frame period is given. The given stream is converted to a frequency-domain dataset X=(X[k].vertline.k=1, 2, . . . N), where k is frequency and N is the total number of subdivided (i.e., discretized) frequency bands.”) (“[0055] It is generally known that speech signals have different spectral envelopes and pitch structures, which result in uneven distribution of frequency components. Spectral envelopes represent the timbre of voice, which is determined by the shape of a speaker's vocal tract (i.e., structure of organs from vocal chords to mouth). A change in the shape of a vocal tract affects its transfer function including resonance characteristics, thus causing uneven distribution of acoustic energies over frequency. Pitch structures indicate the tone height, which comes from the frequency of vocal chord vibration. A temporal change in the pitch structure gives a particular accent or intonation in speech. Background noises, on the other hand, are known to have a relatively uniform spectrum. For this reason, white noise approximation or pink noise approximation is often made to represent them.”) (“[0092] … It sets an appropriate flag to indicate the result. FIG. 15 illustrates how talkspurts are differentiated from noise periods, where the horizontal axis represents frames (time) …”) (“0054] Talkspurt periods can be distinguished from noise periods by calculating the flatness of a power spectrum in the way described above. The following will explain how the spectral flatness varies depending on whether the signal contains speech or only background noise.”) (“[0058] The flatness factor FLT1 of signal X1 (FIG. 7) is obviously greater than FLT2 of signal X2 (FIG. 8). This fact indicates that the signal X1 is speech while the signal X2 is noise. Note here that a larger value of FLT means a less flat spectrum, and that a smaller value of FLT means a flatter spectrum. Talkspurts can be identified by calculating flatness factors of spectrums and comparing them (the voice/noise discriminator 13 actually compares the flatness factor with a predetermined threshold).”) Otani teaches (“[0114] (S14) The voice/noise discriminator 33f compares the flatness factor of each frame with a predetermined threshold. Through this comparison the voice/noise discriminator 33f determines whether the frame in question is speech or noise, and it sets an appropriate flag to indicate the result.”) by Otani et al. US 20050108004 A1.
Otani teaches FFT applies to the input signal, calculate flatness of the spectrum and compare with threshold if speech or noise is present.
Otani is considered to be analogous to the claimed invention because it relates to a voice activity detector, and more particularly to a voice activity detector which discriminates talkspurts from background noises in a given input signal .
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York to incorporate the teachings of Otani in order to include temporal length.
One could have been motivated to do so because model can sets an appropriate flag to indicate the result. (“[0114] … and it sets an appropriate flag to indicate the result.”) by Otani et al. US 20050108004 A1
Claim 19 is a method claim with limitations similar to the limitations of method Claim 1 and is rejected under similar rationale. Additionally:
York teaches:
19. A machine-implemented method for detecting voice activity, the method comprising: receiving a digital representation of an audio signal; York teaches (“(9) … obtain one or more data streams from one or more sensors; determine, based on the one or more data streams, whether a user associated with the electronic device is speaking; in accordance with a determination that the user is speaking: …” col. 2, lines 42-45) (“(218) In some examples, determination module 850 includes speech detector 852 configured to process data streams. In some examples, speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. …” col. 46, lines 56-57 ) by York et al. US 11705130 B2
establishing respective frames defining specified durations within the digital representation of the audio signal; York teaches (“(218) In some examples, determination module 850 includes speech detector 852 configured to process data streams. In some examples, speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. In some examples, speech detector 852 divides an input data stream into frames (e.g., portions) of a predetermined length, e.g., 25, 50, 100, 200, 300, 400, or 500 milliseconds. In some examples, speech detector 852 determines whether each frame indicates that a user is speaking (indicates user speech). For example, speech detector 852 determines a probability of user speech for each frame and/or makes a “yes” or “no” decision regarding the presence of user speech for each frame. In this manner, speech detector 852 can identify the boundaries of user speech in a data stream and determine the duration(s) during which a user is speaking or not speaking.” Col. 46, lines 65-67) (“… For example, speech detector 852 processes the first data stream to determine that a 2 second portion of the data stream indicates user speech. For example, speech detector 852 determines that each frame of a consecutive series of frames (having a collective 2 second duration) indicates user speech.” col. 47, lines 33- 54) by York et al. US 11705130 B2
applying a first stage comprising determining a first frequency-domain indicator from at least one of the respective frames of the digital representation of the audio signal to identify a candidate speech duration; York teaches (“(221) In some examples, determination module 850 determines whether a first data stream obtained from a speech sensor (or a portion thereof) indicates user speech. In some examples, determining whether the first data stream indicates user speech includes analyzing the time domain and/or frequency domain features of the first data stream to determine whether such features indicate human speech. Exemplary time domain features include zero crossing rates, short-time energy, spectral energy, spectral flatness, and autocorrelation. Exemplary frequency domain features include mel-frequency cepstral coefficients, linear predictive cepstral coefficients, and mel-frequency discrete wavelet coefficients. In addition to analyzing such features, one of skill in the art will appreciate that any other suitable technique (e.g., comparing the data stream to a human speech model) may be employed to determine whether a data stream indicates user speech. For example, speech detector 852 processes the first data stream to determine that a 2 second portion of the data stream indicates user speech. For example, speech detector 852 determines that each frame of a consecutive series of frames (having a collective 2 second duration) indicates user speech.” col. 47, lines 33- 54) by York et al. US 11705130 B2
Claim 20 is a system claim with limitations similar to the limitations of method Claim 1 and is rejected under similar rationale. Additionally:
Regarding Claim 20, York teaches
20. A voice activity detection (VAD) system, the system comprising: a receiver circuit, configured to receive a digital representation of an audio signal; and a processor circuit coupled with a memory circuit, the memory circuit containing instructions that, when executed by the processor circuit, cause the processor circuit to: (“(40)… Device 200 includes memory 202 (which optionally includes one or more computer-readable storage mediums), memory controller 222, one or more processing units (CPUs) 220, peripherals interface 218, RF circuitry 208, audio circuitry 210, speaker 211, microphone 213, input/output (I/O) subsystem 206, other input control devices 216, and external port 224. …”col. 8, lines 46-53) (“(46) Peripherals interface 218 is used to couple input and output peripherals of the device to CPU 220 and memory 202. The one or more processors 220 run or execute various software programs and/or sets of instructions stored in memory 202 to perform various functions for device 200 and to process data. In some embodiments, peripherals interface 218, CPU 220, and memory controller 222 are implemented on a single chip, such as chip 204. In some other embodiments, they are implemented on separate chips.” Col.10, lines 42-50) by York et al. US 11705130 B2
Regarding Claim 2, the combination teaches the method claim 1 as identified above.
York further teaches:
2. The method of claim 1, comprising establishing respective frames defining specified durations within the digital representation of the audio signal, wherein at least one of the first stage or the second stage operates on at least one of the respective frames. York teaches (“(218) In some examples, determination module 850 includes speech detector 852 configured to process data streams. In some examples, speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. In some examples, speech detector 852 divides an input data stream into frames (e.g., portions) of a predetermined length, e.g., 25, 50, 100, 200, 300, 400, or 500 milliseconds. In some examples, speech detector 852 determines whether each frame indicates that a user is speaking (indicates user speech). For example, speech detector 852 determines a probability of user speech for each frame and/or makes a “yes” or “no” decision regarding the presence of user speech for each frame. In this manner, speech detector 852 can identify the boundaries of user speech in a data stream and determine the duration(s) during which a user is speaking or not speaking.” Col, 46 lines 65-67 and Col 47, lines 1-13) by York et al. US 11705130 B2
Regarding Claim 3, the combination teaches the method claim 2 as identified above.
York further teaches:
3. The method of claim 2, wherein the receiving a digital representation of an audio signal comprises receiving a streamed representation; and wherein the establishing respective frames includes assigning or receiving the respective frames based on the streamed representation. York teaches (“(218) In some examples, determination module 850 includes speech detector 852 configured to process data streams. In some examples, speech detector 852 includes a voice activity detector (VAD) configured to detect speech in a data stream. In some examples, speech detector 852 divides an input data stream into frames (e.g., portions) of a predetermined length, e.g., 25, 50, 100, 200, 300, 400, or 500 milliseconds. In some examples, speech detector 852 determines whether each frame indicates that a user is speaking (indicates user speech). For example, speech detector 852 determines a probability of user speech for each frame and/or makes a “yes” or “no” decision regarding the presence of user speech for each frame. In this manner, speech detector 852 can identify the boundaries of user speech in a data stream and determine the duration(s) during which a user is speaking or not speaking.” Col, 46 lines 65-67 and Col 47, lines 1-13) by York et al. US 11705130 B2
Regarding Claim 4, the combination teaches the method claim 2 as identified above.
York further teaches:
4. The method of claim 2, wherein the first frequency-domain indicator includes determining a representation of a frequency dispersion of spectral components of the digital representation of the audio signal, the dispersion determined from a frequency domain transform corresponding to one frame. York teaches (“(221) In some examples, determination module 850 determines whether a first data stream obtained from a speech sensor (or a portion thereof) indicates user speech. In some examples, determining whether the first data stream indicates user speech includes analyzing the time domain and/or frequency domain features of the first data stream to determine whether such features indicate human speech. Exemplary time domain features include zero crossing rates, short-time energy, spectral energy, spectral flatness, and autocorrelation. Exemplary frequency domain features include mel-frequency cepstral coefficients, linear predictive cepstral coefficients, and mel-frequency discrete wavelet coefficients. In addition to analyzing such features, one of skill in the art will appreciate that any other suitable technique (e.g., comparing the data stream to a human speech model) may be employed to determine whether a data stream indicates user speech. For example, speech detector 852 processes the first data stream to determine that a 2 second portion of the data stream indicates user speech. For example, speech detector 852 determines that each frame of a consecutive series of frames (having a collective 2 second duration) indicates user speech.” col. 47, lines 33- 54) by York et al. US 11705130 B2
Regarding Claim 11, the combination teaches the method claim 1 as identified above.
York further teaches:
11. The method of claim 1, wherein the second stage comprises an MFC indicator. Teaches York teaches (“(221) In some examples, determination module 850 determines whether a first data stream obtained from a speech sensor (or a portion thereof) indicates user speech. In some examples, determining whether the first data stream indicates user speech includes analyzing the time domain and/or frequency domain features of the first data stream to determine whether such features indicate human speech. Exemplary time domain features include zero crossing rates, short-time energy, spectral energy, spectral flatness, and autocorrelation. Exemplary frequency domain features include mel-frequency cepstral coefficients, linear predictive cepstral coefficients, and mel-frequency discrete wavelet coefficients. In addition to analyzing such features, one of skill in the art will appreciate that any other suitable technique (e.g., comparing the data stream to a human speech model) may be employed to determine whether a data stream indicates user speech. For example, speech detector 852 processes the first data stream to determine that a 2 second portion of the data stream indicates user speech. For example, speech detector 852 determines that each frame of a consecutive series of frames (having a collective 2 second duration) indicates user speech.” col. 47, lines 33- 54) by York et al. US 11705130 B2
Regarding Claim 15, the combination teaches the method claim 1 as identified above.
York further teaches:
15. The method of claim 1, comprising sending a duration determined to contain speech to another system. York teaches (“(316) At block 1130, in accordance with a determination that the user is speaking: at least a portion of the one or more data streams to is provided to an external electronic device (e.g., 800, 900), the portion including data representing a received speech input requesting performance of a task associated with the notification. In some examples, the speech input does not include a trigger phrase for initiating a digital assistant.”) by York et al. US 11705130 B2
Regarding Claim 16, the combination teaches the method claim 15 as identified above.
York further teaches
16. The method of claim 15, wherein the sending the duration determined to contain speech to another system occurs at least partially concurrently with the receiving a digital audio signal corresponding to the duration. (“(316) At block 1130, in accordance with a determination that the user is speaking: at least a portion of the one or more data streams to is provided to an external electronic device (e.g., 800, 900), the portion including data representing a received speech input requesting performance of a task associated with the notification. In some examples, the speech input does not include a trigger phrase for initiating a digital assistant.”) by York et al. US 11705130 B2
Regarding Claim 17, the combination teaches the method claim 1 as identified above.
York does not explicitly teach third stage comprising at least one temporal indicator to assess whether the identified candidate speech duration contains speech.
Otani further teaches:
17. The method of claim 1, comprising applying a third stage comprising at least one temporal indicator to assess whether the identified candidate speech duration contains speech. Otani teaches (“[0023] FIG. 13 shows how the flatness of a given signal is evaluated based on the maximum difference between adjacent spectral components.”) (“[0036] Referring to FIG. 1B, signal segments with a flatter frequency spectrum are regarded as noise, and signal segments with a less flat frequency spectrum are regarded as speech. The voice activity detector 10 of the present invention identifies talkspurts in a given signal accurately by evaluating the flatness of power spectrum of an input signal to determine whether each segment of the signal contains speech or noise.”) by Otani et al. US 20050108004 A1
Otani is considered to be analogous to the claimed invention because it relates to a voice activity detector, and more particularly to a voice activity detector which discriminates talkspurts from background noises in a given input signal .
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York to incorporate the teachings of Otani in order to include temporal length.
One could have been motivated to do so because model can sets an appropriate flag to indicate the result. (“[0114] … and it sets an appropriate flag to indicate the result.”) by Otani et al. US 20050108004 A1
Regarding Claim 18, the combination teaches the method claim 17 as identified above.
Otani further teaches:
18. The method of claim 17, wherein the candidate speech duration is determined not to contain speech if a temporal length of the duration is less than a specified value. Otani teaches (“[0092] … It sets an appropriate flag to indicate the result. FIG. 15 illustrates how talkspurts are differentiated from noise periods, where the horizontal axis represents frames (time) …”) (“0054] Talkspurt periods can be distinguished from noise periods by calculating the flatness of a power spectrum in the way described above. The following will explain how the spectral flatness varies depending on whether the signal contains speech or only background noise.”) (“[0058] The flatness factor FLT1 of signal X1 (FIG. 7) is obviously greater than FLT2 of signal X2 (FIG. 8). This fact indicates that the signal X1 is speech while the signal X2 is noise. Note here that a larger value of FLT means a less flat spectrum, and that a smaller value of FLT means a flatter spectrum. Talkspurts can be identified by calculating flatness factors of spectrums and comparing them (the voice/noise discriminator 13 actually compares the flatness factor with a predetermined threshold).”) Otani teaches (“[0114] (S14) The voice/noise discriminator 33f compares the flatness factor of each frame with a predetermined threshold. Through this comparison the voice/noise discriminator 33f determines whether the frame in question is speech or noise, and it sets an appropriate flag to indicate the result.”) by Otani et al. US 20050108004 A1.
Otani is considered to be analogous to the claimed invention because it relates to a voice activity detector, and more particularly to a voice activity detector which discriminates talkspurts from background noises in a given input signal .
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York to incorporate the teachings of Otani in order to include temporal length.
One could have been motivated to do so because model can sets an appropriate flag to indicate the result. (“[0114] … and it sets an appropriate flag to indicate the result.”) by Otani et al. US 20050108004 A1
Claim 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over York and Otani et al. in view of Xiao et al. US 20150179187 A1.
Regarding Claim 5, the combination teaches the method claim 4 as identified above.
The combination does not explicitly teaches compare frequency spectrum with threshold.
Xiao teaches:
5. The method of claim 4, comprising comparing the determined representation of the dispersion with a first threshold and declaring a candidate speech duration in response to a result of the comparison. Xiao teaches (“[0066] The VAD detection technology for voice segment segmentation may be approximately divided into two steps.”) (“[0067] Step 1: Identify, frame by frame, whether each frame in a voice signal segment is active or non-active. According to a common method in the prior art, activity of each frame is determined by calculating information, such as energy and a frequency spectrum, of each frame and comparing the energy and the frequency spectrum of each frame with a threshold. When the energy and the frequency spectrum of each frame are less than the threshold, the frame is defined to be non-active; otherwise, the frame is defined to be active.”) (“[0077] FIG. 5A and FIG. 5B are schematic diagrams of a voice segment segmentation algorithm according to Embodiment 5 of the present invention. For ease of description, B10 is equivalent to T0 in FIG. 4, B21 is equivalent to T1 in FIG. 4, and a duration [B10, B21] is a voice signal segment. The voice signal segment is detected by means of VAD, and it is determined that voice activity of the following durations [B10, T10], [T11, T20] and [T21, B21] are 0, that is, a status is non-active. Voice activity of durations [T10, T11] and [T20, T21] are 1, that is, a status is active.”) by Xiao et al. US 20150179187 A1
Xiao is considered to be analogous to the claimed invention because it relates to the field of audio technologies, and more specifically, to a voice quality monitoring method and apparatus.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York and Otani to incorporate the teachings of Xiao in order to include feature threshold compare with frame.
One could have been motivated to do so because model can analyze long audio signal by using relatively low costs. (“[0005] In view of this, embodiments of the present invention provide a voice quality monitoring method and apparatus, so as to solve a difficult problem of how to perform proper voice quality monitoring on a relatively long audio signal by using relatively low costs.”) by Xiao et al. US 20150179187 A1
Claim 6 and 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over York, Otani, and Xiao in view of Visser et al. US 20130282373 A1.
Regarding Claim 6, the combination teaches the method claim 5 as identified above.
The combination does not explicitly teaches adjusting the first threshold based upon a central tendency of the first frequency-domain indicator determined using multiple frames.
Visser teaches :
6. The method of claim 5, further comprising adjusting the first threshold based upon a central tendency of the first frequency-domain indicator determined using multiple frames. Visser teaches peak of the frequency range. (“[0206] The bin-wise VAD 2087 may determine voice activity based on the peak information, the bin-wise SNR and the frame-wise voice indicator 2079. For example, the bin-wise VAD 2087 may detect voice activity on a bin-wise basis. More specifically, the bin-wise VAD 2087 may determine which of the peaks indicated by the peak map block/module 2083 are speech peaks. The bin-wise VAD 2087 may generate a bin-wise voice indicator 2089, which may indicate any bins for which voice activity is detected. In particular, the bin-wise voice indicator 2089 may indicate speech peaks and/or non-speech peaks in the transformed audio signal 2071. The peak removal block/module 2090 may remove non-speech peaks.[0207] The bin-wise VAD 2087 may indicate peaks that are associated with speech based on distances between adjacent peaks and temporal continuity. For example, the bin-wise VAD 2087 may indicate small peaks (e.g., peaks that are more than a threshold amount (e.g., 30 dB) below the maximum peak). The bin-wise voice indicator 2089 may indicate these small peaks to the peak removal block/module 2090, which may remove the small peaks from the transformed audio signal 2071. For example, if peaks are determined to be significantly lower (e.g., 30 dB) than a maximum peak, they may not be related to the speech envelope and are thus eliminated. [0208] Additionally, if two peaks are within a certain frequency range (e.g., 90 Hz) and their magnitudes are not much different (e.g., less than 12 dB), the lower one may be indicated as a non-speech peak by the bin-wise VAD 2087 and may be removed by the peak removal block/module 2090. The frequency range may be adjusted depending on speakers. For example, the frequency range may be increased for women or children, who have a relatively higher pitch.[0209] The bin-wise VAD 2087 may also detect temporally isolated peaks (based on the peaks indicated by the peak map block/module 2083, for instance). For example, the bin-wise VAD 2087 may compare peaks from one or more other frames (e.g., previous frame(s) and/or subsequent frame(s)) to peaks in a current frame. For instance, the bin-wise VAD 2087 may detect peaks in a frame that do not have a corresponding peak in a previous frame within a particular range. The range may vary based on the location of the peak. For example, the bin-wise VAD may determine that a peak has a corresponding peak in a previous frame (e.g., that the peak is temporally continuous) if a corresponding peak is found in a previous frame within .+-.1 bin for lower-frequency peaks and within .+-.3 bins for higher-frequency peaks. The bin-wise VAD 2087 may indicate temporally isolated peaks (e.g., peaks in a current frame without corresponding peaks in a previous frame) to the peak removal block/module 2090, which may remove the temporally isolated peaks from the transformed audio signal 2071.”) by Visser et al. US 20130282373 A1
Visser is considered to be analogous to the claimed invention because it relates to systems and methods for audio signal processing.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York, Otani, and Xiao to incorporate the teachings of Visser in order to include feature threshold compare with frequency peak.
One could have been motivated to do so because model can improve voice detection. (“[0061] The techniques disclosed herein may be used to improve voice activity detection (VAD) in order to enhance speech processing, such as voice coding. The disclosed voice activity detection techniques may be used to improve the accuracy and reliability of voice detection, and thus, to improve functions that depend on voice activity detection, ….”) by Visser et al. US 20130282373 A1
Regarding Claim 7, the combination teaches the method claim 6 as identified above.
Visser further teaches :
7. The method of claim 6, wherein the first threshold is adjusted based upon frames that are determined not to contain speech. Visser teaches (“[0208] Additionally, if two peaks are within a certain frequency range (e.g., 90 Hz) and their magnitudes are not much different (e.g., less than 12 dB), the lower one may be indicated as a non-speech peak by the bin-wise VAD 2087 and may be removed by the peak removal block/module 2090. The frequency range may be adjusted depending on speakers. For example, the frequency range may be increased for women or children, who have a relatively higher pitch.”) by Visser et al. US 20130282373 A1
Visser is considered to be analogous to the claimed invention because it relates to systems and methods for audio signal processing.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York, Otani, and Xiao to incorporate the teachings of Visser in order to include feature threshold compare with frequency peak.
One could have been motivated to do so because model can improve voice detection. (“[0061] The techniques disclosed herein may be used to improve voice activity detection (VAD) in order to enhance speech processing, such as voice coding. The disclosed voice activity detection techniques may be used to improve the accuracy and reliability of voice detection, and thus, to improve functions that depend on voice activity detection, ….”) by Visser et al. US 20130282373 A1
Claim 8 and 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over York et al., Otani in view of Lee et al. US 20220108687 A1
Regarding Claim 8, the combination teaches the method claim 2 as identified above.
York does not explicitly teach a logarithm of a frequency domain transform of a time-domain representation.
Lee teaches:
8. The method of claim 2, wherein the second stage comprises a pitch indicator, the pitch indicator comprising an inverse frequency domain transform of: a logarithm of a frequency domain transform of a time-domain representation of a respective one of the frames amongst the respective frames. Lee teaches (“[0093] Next the frequency domain data 136 are transformed by taking the log of the magnitude of the frequency coefficients at 138 and then applying an inverse Fourier transform using FFT 140. These operations transform the data into cepstral coefficients 144 in the frequency domain. Representing the speech signal in the frequency domain is useful in analyzing at what frequencies the sound pressure levels peak in a speech utterance.”) (“[0115] Another subclass of spectrally derived features is known as cepstral based features. The cepstrum is essentially the power spectrum of the logpower spectrum. Given the log power spectrum, as derived above, we can compute the power cepstrum for the kth frame as: … … 1. Cepstral peaks can be used to identify the fundamental frequency, FO, that is, pitch estimation. Cepstral peak:
ceps=DCT(log(|FFT(x)|.sup.2)) …”) (“[0123] In the Short-Term Fourier Transform (STFT) domain, the harmonics of the pitch frequency for voice frames become evident in the magnitude spectrum of the signal. The STFT is formed by taking the DFT from Hamming windowed buffered signal frames with possible zero padding. This observation serves as the basis for the harmonic product spectrum technique which has been utilized for noise-robust pitch detection. The HPS in the log-spectral domain is defined as: …”) by Lee et al. US 20220108687 A1
Lee is considered to be analogous to the claimed invention because it relates to voice detection and speech signal processing.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York, and Otani to incorporate the teachings of Lee in order to include pitch indicator.
One could have been motivated to do so because model can improve recognizability. (“[0003] … Speech enhancement (SE) systems are sometimes used to improve recognizability. …”) by Lee et al. US 20220108687 A1
Regarding Claim 9, The combination teaches the method claim 8 as identified above.
The combination does not explicitly teach wherein the pitch indicator includes determining a central tendency of a magnitude of a specified range of bins within the inverse frequency domain transform.
Lee teaches:
9. The method of claim 8, wherein the pitch indicator includes determining a central tendency of a magnitude of a specified range of bins within the inverse frequency domain transform. Lee teaches (“[0093] Next the frequency domain data 136 are transformed by taking the log of the magnitude of the frequency coefficients at 138 and then applying an inverse Fourier transform using FFT 140. These operations transform the data into cepstral coefficients 144 in the frequency domain. Representing the speech signal in the frequency domain is useful in analyzing at what frequencies the sound pressure levels peak in a speech utterance.”) (“[0115] Another subclass of spectrally derived features is known as cepstral based features. The cepstrum is essentially the power spectrum of the logpower spectrum. Given the log power spectrum, as derived above, we can compute the power cepstrum for the kth frame as: … … 1. Cepstral peaks can be used to identify the fundamental frequency, FO, that is, pitch estimation. Cepstral peak:
ceps=DCT(log(|FFT(x)|.sup.2)) …”) (“[0123] In the Short-Term Fourier Transform (STFT) domain, the harmonics of the pitch frequency for voice frames become evident in the magnitude spectrum of the signal. The STFT is formed by taking the DFT from Hamming windowed buffered signal frames with possible zero padding. This observation serves as the basis for the harmonic product spectrum technique which has been utilized for noise-robust pitch detection. The HPS in the log-spectral domain is defined as: …”) (“[0124] The periodicity is computed as the maximum peak of P(t; !) in the plausible pitch range: …”)m by Lee et al. US 20220108687 A1
Lee is considered to be analogous to the claimed invention because it relates to voice detection and speech signal processing.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York and Otani to incorporate the teachings of Lee in order to include pitch indicator.
One could have been motivated to do so because model can improve recognizability. (“[0003] … Speech enhancement (SE) systems are sometimes used to improve recognizability. …”) by Lee et al. US 20220108687 A1
Claim 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over York, Otani, and Lee in view of Xiao et al. US 20150179187 A1
Regarding Claim 10, The combination teaches the method claim 9 as identified above.
Xiao teaches:
10. The method of claim 9, comprising comparing the determined central tendency to a threshold and declaring a candidate speech duration to be speech if the threshold is exceeded. Fig. 2, Xiao teaches average value (i.e. central tendency) (“[0058] S22. For each input signal of the unit of time, determine whether an average value of the numbers of pitch components included in each input signal of the unit of time is larger. The average value of the numbers of pitch components is compared with a threshold, if the average value of the numbers of pitch components is larger, that is, a result of the determining in S22 is "yes", S23 is performed. Otherwise, a result of the determining in S22 is "no", S24 is performed.”) (“[0060] S24. For each input signal of the unit of time, determine whether a distribution ratio of the pitch components of each input signal of the unit of time at a low frequency is smaller. The distribution ratio of the pitch components is compared with a threshold, if the distribution ratio of the pitch components at a low frequency is smaller, that is, a result of the determining in S24 is "yes", S23 is performed. …”) by Xiao et al. US 20150179187 A1
Xiao is considered to be analogous to the claimed invention because it relates to systems and methods for audio signal processing.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York, Otani and Lee to incorporate the teachings of Xiao in order to include feature threshold compare between spectral centroids.
One could have been motivated to do so because model can analyze long audio signal by using relatively low costs. (“[0005] In view of this, embodiments of the present invention provide a voice quality monitoring method and apparatus, so as to solve a difficult problem of how to perform proper voice quality monitoring on a relatively long audio signal by using relatively low costs.”) by Xiao et al. US 20150179187 A1
Claim 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over York et al. and Otani in view of Visser et al. US 20130282373 A1.
Regarding Claim 12, The combination teaches the method claim 11 as identified above.
The combination does not explicitly wherein the MFC indicator includes determining a representation of a dispersion of the MFC of the digital representation of the audio signal, the dispersion determined from an MFC transform corresponding to at least two frames.
Visser teaches:
12. The method of claim 11, wherein the MFC indicator includes determining a representation of a dispersion of the MFC of the digital representation of the audio signal, the dispersion determined from an MFC transform corresponding to at least two frames. Visser teaches peak of the frequency range.(“[0184] … a noise statistic (e.g., spectral flatness measure) estimation block/module 1747, TF phase voice activity detection/gain difference based suppression block/module 1749, voice activity detection-based residual noise suppression block/module 1751, comb filtering block/module 1755 and an inverse fast Fourier transform block module 1757 that process one or more intermediate signals 1776a-f into an output signal 1780. …”) (“[0206] The bin-wise VAD 2087 may determine voice activity based on the peak information, the bin-wise SNR and the frame-wise voice indicator 2079. For example, the bin-wise VAD 2087 may detect voice activity on a bin-wise basis. More specifically, the bin-wise VAD 2087 may determine which of the peaks indicated by the peak map block/module 2083 are speech peaks. The bin-wise VAD 2087 may generate a bin-wise voice indicator 2089, which may indicate any bins for which voice activity is detected. In particular, the bin-wise voice indicator 2089 may indicate speech peaks and/or non-speech peaks in the transformed audio signal 2071. The peak removal block/module 2090 may remove non-speech peaks.[0207] The bin-wise VAD 2087 may indicate peaks that are associated with speech based on distances between adjacent peaks and temporal continuity. For example, the bin-wise VAD 2087 may indicate small peaks (e.g., peaks that are more than a threshold amount (e.g., 30 dB) below the maximum peak). The bin-wise voice indicator 2089 may indicate these small peaks to the peak removal block/module 2090, which may remove the small peaks from the transformed audio signal 2071. For example, if peaks are determined to be significantly lower (e.g., 30 dB) than a maximum peak, they may not be related to the speech envelope and are thus eliminated. [0208] Additionally, if two peaks are within a certain frequency range (e.g., 90 Hz) and their magnitudes are not much different (e.g., less than 12 dB), the lower one may be indicated as a non-speech peak by the bin-wise VAD 2087 and may be removed by the peak removal block/module 2090. The frequency range may be adjusted depending on speakers. For example, the frequency range may be increased for women or children, who have a relatively higher pitch.[0209] The bin-wise VAD 2087 may also detect temporally isolated peaks (based on the peaks indicated by the peak map block/module 2083, for instance). For example, the bin-wise VAD 2087 may compare peaks from one or more other frames (e.g., previous frame(s) and/or subsequent frame(s)) to peaks in a current frame. For instance, the bin-wise VAD 2087 may detect peaks in a frame that do not have a corresponding peak in a previous frame within a particular range. The range may vary based on the location of the peak. For example, the bin-wise VAD may determine that a peak has a corresponding peak in a previous frame (e.g., that the peak is temporally continuous) if a corresponding peak is found in a previous frame within .+-.1 bin for lower-frequency peaks and within .+-.3 bins for higher-frequency peaks. The bin-wise VAD 2087 may indicate temporally isolated peaks (e.g., peaks in a current frame without corresponding peaks in a previous frame) to the peak removal block/module 2090, which may remove the temporally isolated peaks from the transformed audio signal 2071.” [0295].. The term "frequency component" is used to indicate one among a set of frequencies or frequency bands of a signal, such as a sample of a frequency domain representation of the signal (e.g., as produced by a fast Fourier transform) or a subband of the signal (e.g., a Bark scale or mel scale subband) by Visser et al. US 20130282373 A1
Visser is considered to be analogous to the claimed invention because it relates to systems and methods for audio signal processing.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York, and Otani to incorporate the teachings of Visser in order to include feature threshold compare with frequency peak.
One could have been motivated to do so because model can improve voice detection. (“[0061] The techniques disclosed herein may be used to improve voice activity detection (VAD) in order to enhance speech processing, such as voice coding. The disclosed voice activity detection techniques may be used to improve the accuracy and reliability of voice detection, and thus, to improve functions that depend on voice activity detection, ….”) by Visser et al. US 20130282373 A1.
Claim 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over York, Otani and Visser et al. US 20130282373 A1 in view of Xiao, CN 104409081 A
Regarding Claim 13, The combination teaches the method claim 12 as identified above.
The combination does not explicitly compare the determined representation of dispersion of the MFC to at least one threshold and at least one of adjusting a candidate speech duration or declaring a candidate speech duration to be speech in response to a result of the comparison.
Xiao teaches:
13. The method of claim 12, comprising comparing the determined representation of dispersion of the MFC to at least one threshold and at least one of adjusting a candidate speech duration or declaring a candidate speech duration to be speech in response to a result of the comparison. Xiao teaches (“[0151] When the second characteristic value is the cepstrum distance, detecting the cepstrum distance of the voice signal is greater than a preset cepstrum distance threshold value, if it is more than the preset cepstrum distance threshold, it is determined that the voice signal belonging to unvoiced signal. otherwise, determining that the voice signal belonging to a voice signal. wherein the preset cepstrum distance threshold value is preset according to the actual requirement of the empirical value.”) (“[0220] when the first characteristic value is a cepstrum distance, detecting the cepstrum distance of the voice signal is greater than a preset cepstrum distance threshold value, if it is more than the preset cepstrum distance threshold value, determining the frame voice signal to voice signal. otherwise, determining that the frame voice signal to the voice signal. wherein the preset cepstrum distance threshold value is preset according to the actual requirement of the empirical value.”)
Xiao is considered to be analogous to the claimed invention because it relates to generally to the field of speech recognition.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York, Otani, and Visser to incorporate the teachings of Xiao in order to include feature threshold compare with frame.
One could have been motivated to do so because model improves the real-time performance of the beatbox treatment and efficiency. (“[0092] … low efficiency problem, improves the real-time performance of the beatbox treatment and efficiency, and there is no need for artificial post repair, reaches the effect of automatically detecting beatbox. …”) by Xiao, CN 104409081 A
Claim 14, is/are rejected under 35 U.S.C. 103 as being unpatentable over York, Otani in view of JANKOWSKI et al. US 20200074997 A1
Regarding Claim 14, York teaches the method claim 1 as identified above.
York does not explicitly teach MFC indicator and a pitch indicator.
JANKOWSKI teaches:
14. The method of claim 1, wherein the second stage comprises both an MFC indicator and a pitch indicator. JANKOWSKI teaches (“A voice activity detection method includes: training one or more computerized neural networks having a denoising autoencoder and a classifier, wherein the training is performed utilizing one or more models including Mel-frequency cepstral coefficients (MFCC) features, Δ features, ΔΔ features, and Pitch features, …”) by JANKOWSKI et al. US 20200074997 A1
JANKOWSKI is considered to be analogous to the claimed invention because it relates to voice recognition systems and methods for extracting speech and filtering speech from other audio waveforms.
Therefore, it would have been obvious for someone of ordinary skill in the art before the effective filing date of the claimed invention to modify the York and Otani to incorporate the teachings of JANKOWSKI in order to include both an MFC indicator and a pitch indicator.
One could have been motivated to do so because model add robustness to voice activity detection in noisy conditions. (“[0066] Various embodiments of the present disclosure provide improvements over existing VAD systems by utilizing a series of techniques that add robustness to voice activity detection in noisy conditions, …”) by JANKOWSKI et al. US 20200074997 A1
Conclusion
THIS ACTION IS MADE FINAL. 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 FOUZIA HYE SOLAIMAN whose telephone number is (571)270-5656. The examiner can normally be reached M-F (8-5)AM.
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, Paras D. Shah can be reached at (571) 270-1650. 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.
/F.H.S./Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
07/17/2026
Prosecution Insights