DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The amendments filed February 10th, 2026 have been entered. Claims 1-3, 7, and 12-15 have been amended. Claims 1-20 are pending and have been examined. Applicant’s amendments have overcome all objections to the claims due to informalities previously set forth.
Claims 2 and 14 were previously objected to, but were deemed to contain allowable subject matter. However, in view of the newly submitted IDS documents, and necessitated by the amendments to independent claims 1 and 14, claims 2 and 14 have been rejected under 35 U.S.C. 103 as described below.
Response to Arguments
Applicant’s arguments with respect to claim(s) 1, 4-13, and 16-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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(s) 1-2, 4-5, 10, 12-14, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (US Pat. Pub. No. 2022/0301573 A1 hereinafter Wang), in view of Al-Hussaini et al. (Al-Hussaini, E.K., & Ahsanullah, M. (2015). Exponentiated distributions. Atlantis Press. 4 pages. hereinafter Al-Hussaini) and Borgstrom et al. (US Pat. Pub. No. 2021/0074282 A1 hereinafter Borgstrom).
Regarding claim 1, Wang discloses a computer-implemented method of mitigating over-suppression of speech, comprising: receiving, by a processor, audio data (Wang, [0030]: "In some implementations, the voice filter model can operate as a frame-by-frame frontend signal processor to enhance the features consumed by a speech recognizer, without reconstructing audio signals from the features. During training, an asymmetric loss function can be utilized to penalize over-suppression which can make the model harmless under more acoustic environments."); executing a digital model for detecting speech on features of the audio data, the digital model being trained with a loss function with non-linear penalty that penalizes speech over-suppression more than speech under-suppression (Wang, [0012]: "Utilizing an asymmetric loss when training the voice filter model in place of a conventional loss function can result in a trained model which gives more tolerance to under suppression errors and less tolerance to over suppression errors."), the digital model configured to produce a mask of estimated mask values indicating an amount of speech present for each frame of the plurality of frames and each frequency band of the plurality of frequency bands ([0003]: "Techniques described herein are directed to isolating a human voice from a frequency representation of an audio signal by generating a predicted mask using a trained voice filter model, wherein the frequency representation is generated using an automatic speech recognition (ASR) engine, and where processing the frequency representation with the predicted mask can isolate portion(s) of the frequency representation corresponding to the human voice."). However, Wang fails to expressly recite the non-linear penalty comprising an exponential function of a difference between a target mask value and an estimated mask value; audio data as a joint time-frequency representation over a plurality of frames and a plurality of frequency bands; and transmitting information regarding the mask to a device.
Al-Hussaini teaches the non-linear penalty comprising an exponential function of a difference between a target mask value and an estimated mask value (Al-Hussaini, page 6, equation 1.3.3, discloses a non-linear loss penalty comprising an exponential function.).
PNG
media_image1.png
35
421
media_image1.png
Greyscale
Al-Hussaini, equation 1.3.3 for reference
Wang and Al-Hussaini are analogous arts because they both belong to the field of data processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang to incorporate the teachings of Al-Hussaini to use a non-linear, exponential loss function. A function of this type can unequally effect under-estimation and over-estimation since one may be more serious than the other (Al-Hussaini, page 6). This ensures that the system can avoid over-suppression of speech. However, Wang, in view of Al-Hussaini, fails to expressly recite audio data as a joint time-frequency representation over a plurality of frames and a plurality of frequency bands; and transmitting information regarding the mask to a device.
Borgstrom teaches audio data as a joint time-frequency representation over a plurality of frames and a plurality of frequency bands (Borgstrom, [0011]: "The deep neural network can be configured to predict voice activity (e.g., the probability of the presence of speech) on a per-frame and per-frequency-band basis of an input audio spectrum."); and transmitting information regarding the mask to a device (Borgstrom, [0074]: "In some implementations, the input/output device 1040 can include driver devices configured to receive input data and send output data to other input/output devices, e.g., a keyboard, a printer, and display devices (such as the GUI 12).").
Wang, Al-Hussaini, and Borgstrom are analogous arts because they each belong to the field of data processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini, to incorporate the teachings of Borgstrom to process combined time and frequency audio data over a plurality of frames and frequency bands. This helps the system provide significant levels of noise suppression while maintaining high speech quality (Borgstrom, [0021]). These benefits are important for providing high quality audio to other processes as well as to users of the system.
Regarding claim 2, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. Al-Hussaini further teaches the loss function being mdiff-diff-1,and wherein diff denotes a difference between the target mask value with a power-law term and the estimated mask value of the estimated mask values with the power-law term, and m denotes a tuning parameter (Al-Hussaini, page 6, equation 1.3.3, discloses a non-linear loss penalty comprising an exponential function of comparable form to the claimed equation.). The same motivation for claim 1 applies equally to claim 2.
Regarding claim 4, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. Borgstrom further teaches the joint time-frequency representation having an energy value for each time frame and each frequency band, the method further comprising computing a logarithm of each energy value in the joint time-frequency representation as a feature of the features (Borgstrom, [0011]: "The deep neural network can be configured to predict voice activity (e.g., the probability of the presence of speech) on a per-frame and per-frequency-band basis of an input audio spectrum."; [0047]: "The features used for detection of speech presence can be based on log-spectra and modified for robustness"). The same motivation for claim 1 applies equally to claim 4.
Regarding claim 5, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. Borgstrom further teaches the digital model being an artificial neural network trained using a training dataset of joint time-frequency representations of different mixtures of speech and non-speech (Borgstrom, [0011]: "The deep neural network can be trained using speech conversations created using speech data containing noise and/or distortion, and, in at least some instances, the speech data can be created using a combination of clean speech data and silence data."). The same motivation for claim 1 applies equally to claim 5.
Regarding claim 10, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. Borgstrom further teaches receiving an input waveform in a time domain (Borgstrom, [0045]: "the input processor 90 can generate, from a time-domain signal 101, a series of frames of pseudo-stationary segments and then apply the Fourier transform of each segment to generate the local spectra for each frame."); transforming the input waveform into raw audio data over a plurality of frequency bins and the plurality of frames; and converting the raw audio data into the audio data by grouping the plurality of frequency bins into the plurality of frequency bands (Borgstrom, [0045]: "The DNN 120 can be trained to make binary classifications for each spectrogram bin, where a bin is an individual frequency band of a single frame. For example, the DNN 120 can be trained to output 0 if the bin does not contain active speech with 100% probability, and 1 if the bin contains active speech with 100% probability, and any value between 0 and 1 to represent probabilities between 0 and 100%."). The same motivation for claim 1 applies equally to claim 10.
Regarding claim 12, Wang discloses a system for mitigating over-suppression of speech, comprising: a memory; and one or more processors coupled to the memory and configured to perform (Wang, [0016]: “some implementations include one or more processors (e.g., central processing unit(s) (CPU(s)), graphics processing unit(s) (GPU(s), and/or tensor processing unit(s) (TPU(s)) of one or more computing devices, where the one or more processors are operable to execute instructions stored in associated memory, and where the instructions are configured to cause performance of any of the methods described herein.”): receiving, by a processor, audio data (Wang, [0030]: "In some implementations, the voice filter model can operate as a frame-by-frame frontend signal processor to enhance the features consumed by a speech recognizer, without reconstructing audio signals from the features. During training, an asymmetric loss function can be utilized to penalize over-suppression which can make the model harmless under more acoustic environments."); executing a digital model for detecting speech on features of the audio data, the digital model being trained with a loss function with non-linear penalty that penalizes speech over-suppression more than speech under-suppression (Wang, [0012]: "Utilizing an asymmetric loss when training the voice filter model in place of a conventional loss function can result in a trained model which gives more tolerance to under suppression errors and less tolerance to over suppression errors."), the digital model configured to produce a mask of estimated mask values indicating an amount of speech present for each frame of the plurality of frames and each frequency band of the plurality of frequency bands ([0003]: "Techniques described herein are directed to isolating a human voice from a frequency representation of an audio signal by generating a predicted mask using a trained voice filter model, wherein the frequency representation is generated using an automatic speech recognition (ASR) engine, and where processing the frequency representation with the predicted mask can isolate portion(s) of the frequency representation corresponding to the human voice."). However, Wang fails to expressly recite the non-linear penalty comprising an exponential function of a difference between a target mask value and an estimated mask value; audio data as a joint time-frequency representation over a plurality of frames and a plurality of frequency bands; and transmitting information regarding the mask to a device.
Al-Hussaini teaches the non-linear penalty comprising an exponential function of a difference between a target mask value and an estimated mask value (Al-Hussaini, page 6, equation 1.3.3, discloses a non-linear loss penalty comprising an exponential function.).
Wang and Al-Hussaini are analogous arts because they both belong to the field of data processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang to incorporate the teachings of Al-Hussaini to use a non-linear, exponential loss function. A function of this type can unequally effect under-estimation and over-estimation since one may be more serious than the other (Al-Hussaini, page 6). This ensures that the system can avoid over-suppression of speech. However, Wang, in view of Al-Hussaini, fails to expressly recite audio data as a joint time-frequency representation over a plurality of frames and a plurality of frequency bands; and transmitting information regarding the mask to a device.
Borgstrom teaches audio data as a joint time-frequency representation over a plurality of frames and a plurality of frequency bands (Borgstrom, [0011]: "The deep neural network can be configured to predict voice activity (e.g., the probability of the presence of speech) on a per-frame and per-frequency-band basis of an input audio spectrum."); and transmitting information regarding the mask to a device (Borgstrom, [0074]: "In some implementations, the input/output device 1040 can include driver devices configured to receive input data and send output data to other input/output devices, e.g., a keyboard, a printer, and display devices (such as the GUI 12).").
Wang, Al-Hussaini, and Borgstrom are analogous arts because they each belong to the field of data processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini, to incorporate the teachings of Borgstrom to process combined time and frequency audio data over a plurality of frames and frequency bands. This helps the system provide significant levels of noise suppression while maintaining high speech quality (Borgstrom, [0021]). These benefits are important for providing high quality audio to other processes as well as to users of the system.
Regarding claim 13, Wang discloses a computer-readable, non-transitory storage medium storing computer-executable instructions (Wang, [0016]: “Some implementations also include one or more non-transitory computer readable storage media storing computer instructions executable by one or more processors to perform any of the methods described herein.”), which when executed implement a method of mitigating over-suppression of speech, the method comprising: creating a digital model for detecting speech from the training dataset using a loss function with non-linear penalty that penalizes speech over-suppression more than speech under-suppression (Wang, [0012]: "Utilizing an asymmetric loss when training the voice filter model in place of a conventional loss function can result in a trained model which gives more tolerance to under suppression errors and less tolerance to over suppression errors."), the digital model configured to produce a mask for in audio data over a plurality of frequency bands and a plurality of frames, the mask including one estimated mask value indicating an amount of detected speech in each frequency band of the plurality of frequency bands at each frame of the plurality of frames ([0003]: "Techniques described herein are directed to isolating a human voice from a frequency representation of an audio signal by generating a predicted mask using a trained voice filter model, wherein the frequency representation is generated using an automatic speech recognition (ASR) engine, and where processing the frequency representation with the predicted mask can isolate portion(s) of the frequency representation corresponding to the human voice."); receiving new audio data; executing a digital model for detecting speech on features of the new audio data to obtain a new mask (Wang, [0030]: "In some implementations, the voice filter model can operate as a frame-by-frame frontend signal processor to enhance the features consumed by a speech recognizer, without reconstructing audio signals from the features. During training, an asymmetric loss function can be utilized to penalize over-suppression which can make the model harmless under more acoustic environments."). However, Wang does not expressly recite the non-linear penalty comprising an exponential function of a difference between a target mask value and an estimated mask value; receiving, by a processor, a training dataset of a plurality of joint time-frequency representations; and transmitting information regarding the new mask to a device.
Al-Hussaini teaches the non-linear penalty comprising an exponential function of a difference between a target mask value and an estimated mask value (Al-Hussaini, page 6, equation 1.3.3, discloses a non-linear loss penalty comprising an exponential function.).
Wang and Al-Hussaini are analogous arts because they both belong to the field of data processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang to incorporate the teachings of Al-Hussaini to use a non-linear, exponential loss function. A function of this type can unequally effect under-estimation and over-estimation since one may be more serious than the other (Al-Hussaini, page 6). This ensures that the system can avoid over-suppression of speech. However, Wang, in view of Al-Hussaini, fails to expressly recite receiving, by a processor, a training dataset of a plurality of joint time-frequency representations; and transmitting information regarding the new mask to a device.
Borgstrom teaches receiving, by a processor, a training dataset of a plurality of joint time-frequency representations (Borgstrom, [0011]: "The deep neural network can be trained using speech conversations created using speech data containing noise and/or distortion, and, in at least some instances, the speech data can be created using a combination of clean speech data and silence data."); and transmitting information regarding the new mask to a device (Borgstrom, [0074]: "In some implementations, the input/output device 1040 can include driver devices configured to receive input data and send output data to other input/output devices, e.g., a keyboard, a printer, and display devices (such as the GUI 12).").
Wang, Al-Hussaini, and Borgstrom are analogous arts because they each belong to the field of data processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini, to incorporate the teachings of Borgstrom to process combined time and frequency audio data over a plurality of frames and frequency bands. This helps the system provide significant levels of noise suppression while maintaining high speech quality (Borgstrom, [0021]). These benefits are important for providing high quality audio to other processes as well as to users of the system.
Regarding claim 14, the rejection of claim 13 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. Al-Hussaini further teaches the loss function being mdiff-diff-1,and wherein diff denotes a difference between the target mask value with a power-law term and the estimated mask value of the estimated mask values with the power-law term, and m denotes a tuning parameter (Al-Hussaini, page 6, equation 1.3.3, discloses a non-linear loss penalty comprising an exponential function of comparable form to the claimed equation.). The same motivation for claim 13 applies equally to claim 14.
Regarding claim 19, the rejection of claim 13 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. Borgstrom further teaches receiving an input waveform in a time domain (Borgstrom, [0045]: "the input processor 90 can generate, from a time-domain signal 101, a series of frames of pseudo-stationary segments and then apply the Fourier transform of each segment to generate the local spectra for each frame."); transforming the input waveform into raw audio data over a plurality of frequency bins and the plurality of frames; and converting the raw audio data into the audio data by grouping the plurality of frequency bins into the plurality of frequency bands (Borgstrom, [0045]: "The DNN 120 can be trained to make binary classifications for each spectrogram bin, where a bin is an individual frequency band of a single frame. For example, the DNN 120 can be trained to output 0 if the bin does not contain active speech with 100% probability, and 1 if the bin contains active speech with 100% probability, and any value between 0 and 1 to represent probabilities between 0 and 100%."). The same motivation for claim 13 applies equally to claim 19.
Claim(s) 6 and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Al-Hussaini and Borgstrom, as applied to claims 1-2, 4-5, 10, 12-14, and 19 above, and further in view of Pereg et al. (US Pat. Pub. No. 2008/0195385 A1 hereinafter Pereg).
Regarding claim 6, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite determining whether the audio data corresponds to laughter or applause; and in response to determining that the audio data corresponds to laughter or applause, further transmitting an alert to ignore the mask.
Pereg teaches determining whether the audio data corresponds to laughter or applause; and in response to determining that the audio data corresponds to laughter or applause, further transmitting an alert to ignore the mask (Pereg, [0019]: "The disclosed invention presents an effective and efficient laughter detection method and apparatus in audio interactions. The method is based on detecting laughter episodes, comprising of at least a minimal predetermined number of consecutive bursts, wherein each burst is composed of a voice portion immediately, or close to immediately, followed by an unvoiced or silent portion. Once a sequence of bursts is identified, laughter characteristic features are determined for the sequence, and are compared against one or more predetermined sets of criteria. If the features meet any of the criteria sets, a predetermined laughter certainty score is attached to the relevant part of the interaction."; Here, since Pereg detects laughter but does not filter it out, it is seen as obvious to ignore the mask when used to modify Wang, in view of Borgstrom.).
Wang, Al-Hussaini, Borgstrom, and Pereg are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Pereg to detect laughter and not apply the mask to the laughter. Detecting laughter independent of speech is important for emotion detection (Pereg, [0004]). As such, detecting and not applying a mask to laughter can enable emotion detection in the output audio.
Regarding claim 16, the rejection of claim 13 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite determining whether the audio data corresponds to laughter or applause; and in response to determining that the audio data corresponds to laughter or applause, further transmitting an alert to ignore the mask.
Pereg teaches determining whether the audio data corresponds to laughter or applause; and in response to determining that the audio data corresponds to laughter or applause, further transmitting an alert to ignore the mask (Pereg, [0019]: "The disclosed invention presents an effective and efficient laughter detection method and apparatus in audio interactions. The method is based on detecting laughter episodes, comprising of at least a minimal predetermined number of consecutive bursts, wherein each burst is composed of a voice portion immediately, or close to immediately, followed by an unvoiced or silent portion. Once a sequence of bursts is identified, laughter characteristic features are determined for the sequence, and are compared against one or more predetermined sets of criteria. If the features meet any of the criteria sets, a predetermined laughter certainty score is attached to the relevant part of the interaction." ; Here, since Pereg detects laughter but does not filter it out, it is seen as obvious to ignore the mask when used to modify Wang, in view of Borgstrom.).
Wang, Al-Hussaini, Borgstrom, and Pereg are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Pereg to detect laughter and not apply the mask to the laughter. Detecting laughter independent of speech is important for emotion detection (Pereg, [0004]). As such, detecting and not applying a mask to laughter can enable emotion detection in the output audio.
Claim(s) 7 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Al-Hussaini and Borgstrom, as applied to claims 1-2, 4-5, 10, 12-14, and 19 above, and further in view of Bastian et al. (US Pat. Pub. No. 2018/0336920 A1 hereinafter Bastian).
Regarding claim 7, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite computing derived features of the audio data in a time domain and a frequency domain; and executing a second digital model for classifying the audio data into laughter or applause or otherwise based on the derived features.
Bastian teaches computing derived features of the audio data in a time domain and a frequency domain; and executing a second digital model for classifying the audio data into laughter or applause or otherwise based on the derived features (Bastian, [0079]: "changes over time of the frequency domain data of the transformed speech data received at 202 are used to project (predict) corresponding future changes in the received speech signal of the speaker, to thereby predict which of the bands of frequencies transformed from the analog speech signal data of the speaker will have peak values in the future. As the background noise frequency values tend to remain steady or constant over time, this process identifies attributes of the speech signals generated by the speaker that represent anomalies, such as excited utterances, laughter and other emphatic and non-verbal signals that tend to have frequency profiles more pronounced and different from speech signals generated from word content. The remainder of the speech signal data comprises the background noise along with an average signal of speech data generated by the speaker, wherein the speaker signal data is smoothed out by removing the forecasted peaks.").
Wang, Al-Hussaini, Borgstrom, and Bastian are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Bastian to use a digital model to detect classify audio based on features. This allows anomalies in an audio signal to be detected and identified (Bastian, [0079]). Detecting and identifying anomalies ensures the system can effectively process different types of signals with more than just speech and general background noise.
Regarding claim 17, the rejection of claim 13 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite computing derived features of the audio data in a time domain and a frequency domain; and executing a second digital model for classifying the audio data into laughter or applause or otherwise based on the derived features.
Bastion teaches computing derived features of the audio data in a time domain and a frequency domain; and executing a second digital model for classifying the audio data into laughter or applause or otherwise based on the derived features (Bastian, [0079]: "changes over time of the frequency domain data of the transformed speech data received at 202 are used to project (predict) corresponding future changes in the received speech signal of the speaker, to thereby predict which of the bands of frequencies transformed from the analog speech signal data of the speaker will have peak values in the future. As the background noise frequency values tend to remain steady or constant over time, this process identifies attributes of the speech signals generated by the speaker that represent anomalies, such as excited utterances, laughter and other emphatic and non-verbal signals that tend to have frequency profiles more pronounced and different from speech signals generated from word content. The remainder of the speech signal data comprises the background noise along with an average signal of speech data generated by the speaker, wherein the speaker signal data is smoothed out by removing the forecasted peaks.").
Wang, Al-Hussaini, Borgstrom, and Bastian are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Bastian to use a digital model to detect classify audio based on features. This allows anomalies in an audio signal to be detected and identified (Bastian, [0079]). Detecting and identifying anomalies ensures the system can effectively process different types of signals with more than just speech and general background noise.
Claim(s) 8 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Al-Hussaini and Borgstrom, as applied to claims 1-2, 4-5, 10, 12-14, and 19 above, and further in view of Suzuki et al. (US Pat. Pub. No. 2005/0165608 A1 hereinafter Suzuki).
Regarding claim 8, the rejection of claim 1 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite computing a mask attenuation for the mask; determining whether the mask attenuation corresponds to a fall-off amount that exceeds a threshold; and in response to determining that the mask attenuation corresponds to a fall-off amount that exceeds the threshold, adjusting the mask such that the mask attenuation matches a predetermined voice decay rate.
Suzuki teaches computing a mask attenuation for the mask (Suzuki, [0020]: " in cases where the input voice is processed for each frame, the conditions of enhancement (amplification factor or attenuation factor) vary between frames. Accordingly, if the amplification factor or attenuation factor varies abruptly between frames, the feeling of noise is increased by the fluctuation of the spectrum."); determining whether the mask attenuation corresponds to a fall-off amount that exceeds a threshold; and in response to determining that the mask attenuation corresponds to a fall-off amount that exceeds the threshold, adjusting the mask such that the mask attenuation matches a predetermined voice decay rate (Suzuki, [0027]: "a difference calculating part which calculates the difference amplification factor from the abovementioned tentative amplification factor and the amplification factor of the preceding frame, and an amplification factor judgment part which takes the amplification factor determined from a predetermined threshold value and the amplification factor of the preceding frame as the amplification factor of the current frame in cases where the abovementioned difference is greater than this threshold value, and which takes the abovementioned tentative amplification factor as the amplification factor of the current frame in cases where the abovementioned difference is smaller than the abovementioned threshold value").
Wang, Al-Hussaini, Borgstrom, and Suzuki are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Suzuki to detect large shifts in speech signal strength. Abrupt changes in amplification factor or attenuation factor between frames can make the signal seem noisier (Suzuki, [0020]). This should be avoided to ensure the output audio is high quality.
Regarding claim 18, the rejection of claim 13 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite computing a mask attenuation for the mask; determining whether the mask attenuation corresponds to a fall-off amount that exceeds a threshold; and in response to determining that the mask attenuation corresponds to a fall-off amount that exceeds the threshold, adjusting the mask such that the mask attenuation matches a predetermined voice decay rate.
Suzuki teaches computing a mask attenuation for the mask (Suzuki, [0020]: " in cases where the input voice is processed for each frame, the conditions of enhancement (amplification factor or attenuation factor) vary between frames. Accordingly, if the amplification factor or attenuation factor varies abruptly between frames, the feeling of noise is increased by the fluctuation of the spectrum."); determining whether the mask attenuation corresponds to a fall-off amount that exceeds a threshold; and in response to determining that the mask attenuation corresponds to a fall-off amount that exceeds the threshold, adjusting the mask such that the mask attenuation matches a predetermined voice decay rate (Suzuki, [0027]: "a difference calculating part which calculates the difference amplification factor from the abovementioned tentative amplification factor and the amplification factor of the preceding frame, and an amplification factor judgment part which takes the amplification factor determined from a predetermined threshold value and the amplification factor of the preceding frame as the amplification factor of the current frame in cases where the abovementioned difference is greater than this threshold value, and which takes the abovementioned tentative amplification factor as the amplification factor of the current frame in cases where the abovementioned difference is smaller than the abovementioned threshold value").
Wang, Al-Hussaini, Borgstrom, and Suzuki are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Suzuki to detect large shifts in speech signal strength. Abrupt changes in amplification factor or attenuation factor between frames can make the signal seem noisier (Suzuki, [0020]). This should be avoided to ensure the output audio is high quality.
Claim(s) 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Al-Hussaini, Borgstrom, and Suzuki, as applied to claims 8 and 18 above, and further in view of Shi et al. (US Pat. No. 2018/0172502 A1 hereinafter Shi).
Regarding claim 9, the rejection of claim 8 is incorporated. Wang, in view of Al-Hussaini, Borgstrom, and Suzuki, discloses all of the elements of the current invention as stated above. However, Wang, in view of Borgstrom Suzuki, fails to expressly recite the predetermined voice decay rate being 200 ms reverberation time.
Shi teaches the predetermined voice decay rate being 200 ms reverberation time (Shi, [0071]: " it may be assumed that the clean voice power spectrum is largely impulsive, with rapid onset and a decay rate much greater than that for the reverberation. For example, natural voice characteristics are decaying at least 20 or 30 dB within 100 ms, being around half of the normal syllable duration. This would correspond to a room with the reverberation time being less than 200 ms.").
Wang, Al-Hussaini, Borgstrom, Suzuki, and Shi are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini, the speech enhancement improvement methods of Borgstrom, and the voice enhancement device of Suzuki, to incorporate the teachings of Shi to use a predetermined voice decay rate of 200 ms. This is the decay rate of natural voice characteristics in a room (Shi, [0071]). Using this as a baseline rate allows the voice decay rate to sound relatively similar to that of a natural voice by default.
Claim(s) 11 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Al-Hussaini and Borgstrom, as applied to claims 1-2, 4-5, 10, 12-14, and 19 above, and further in view of Goesnar et al. (US Pat. Pub. No. 2016/0035367 A1 hereinafter Goesnar).
Regarding claim 11, the rejection of claim 10 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite performing inverse banding on the estimated mask values to generate updated mask values for each frequency bin of the plurality of frequency bins and each frame of the plurality of frames; applying the updated mask values to the raw audio data to generate new output data; and transforming the new output data into an enhanced waveform.
Goesnar teaches performing inverse banding on the estimated mask values to generate updated mask values for each frequency bin of the plurality of frequency bins and each frame of the plurality of frames (Goesnar, [0095]: "In this implementation, the gains will ultimately be applied to the frequency domain audio data of the M subbands output by the analysis filterbank 1305. Therefore, in this example the inverse banding block 1340 is configured to receive the smoothed gain values for each of the N subbands that are output from the regularization block 1335 and to output smoothed gain values for M subbands."); applying the updated mask values to the raw audio data to generate new output data (Goesnar, [0095]: "Here, the gain applying modules 1345 are configured to apply the smoothed gain values, output by the inverse banding block 1340, to the frequency domain audio data of the M subbands that are output by the analysis filterbank 1305."); and transforming the new output data into an enhanced waveform (Goesnar, [0095]: "Here, the synthesis filterbank 1310 is configured to reconstruct the audio data of the M frequency subbands, with gain values modified by the gain applying modules 1345, into the output signal y[n].").
Wang, Al-Hussaini, Borgstrom, and Goesnar are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Goesnar to perform inverse banding on the mask values, then applying them to the raw audio data, and finally transforming the audio data into a waveform. This allows the system to apply gain values to the audio data and transform the audio data back into an output signal (Goesnar, [0095]). This ensures that the data is output in a usable format despite any transformations that previously took place.
Regarding claim 20, the rejection of claim 19 is incorporated. Wang, in view of Al-Hussaini and Borgstrom, discloses all of the elements of the current invention as stated above. However, Wang, in view of Al-Hussaini and Borgstrom, fails to expressly recite performing inverse banding on the estimated mask values to generate updated mask values for each frequency bin of the plurality of frequency bins and each frame of the plurality of frames; applying the updated mask values to the raw audio data to generate new output data; and transforming the new output data into an enhanced waveform.
Goesnar teaches performing inverse banding on the estimated mask values to generate updated mask values for each frequency bin of the plurality of frequency bins and each frame of the plurality of frames (Goesnar, [0095]: "In this implementation, the gains will ultimately be applied to the frequency domain audio data of the M subbands output by the analysis filterbank 1305. Therefore, in this example the inverse banding block 1340 is configured to receive the smoothed gain values for each of the N subbands that are output from the regularization block 1335 and to output smoothed gain values for M subbands."); applying the updated mask values to the raw audio data to generate new output data (Goesnar, [0095]: "Here, the gain applying modules 1345 are configured to apply the smoothed gain values, output by the inverse banding block 1340, to the frequency domain audio data of the M subbands that are output by the analysis filterbank 1305."); and transforming the new output data into an enhanced waveform (Goesnar, [0095]: "Here, the synthesis filterbank 1310 is configured to reconstruct the audio data of the M frequency subbands, with gain values modified by the gain applying modules 1345, into the output signal y[n].").
Wang, Al-Hussaini, Borgstrom, and Goesnar are analogous arts because they all belong to the field of audio processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the voice separation method of Wang, as modified by the exponential distributions of Al-Hussaini and the speech enhancement improvement methods of Borgstrom, to incorporate the teachings of Goesnar to perform inverse banding on the mask values, then applying them to the raw audio data, and finally transforming the audio data into a waveform. This allows the system to apply gain values to the audio data and transform the audio data back into an output signal (Goesnar, [0095]). This ensures that the data is output in a usable format despite any transformations that previously took place.
Allowable Subject Matter
Claims 3 and 15 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.
The following is a statement of reasons for the indication of allowable subject matter: While the cited prior art discloses an asymmetrical loss function, the prior art search did not yield any references that disclosed the specific loss functions claimed.
Conclusion
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 TYLER J BECKER whose telephone number is (703)756-1271. The examiner can normally be reached M-Th, 7:15am-5:45pm PT.
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, Daniel Washburn can be reached at (571) 272-5551. 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.
/TYLER BECKER/ Examiner, Art Unit 2657
/DANIEL C WASHBURN/ Supervisory Patent Examiner, Art Unit 2657
Prosecution Insights