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 .
DETAILED ACTION
Claims 1, 3-19, 21, and 24 are pending. Claims 1, 21, and 24 are independent.
This Application was published as US 20230326475.
Apparent priority is 6 April 2022.
Response to Arguments
35 USC 101
Applicant's arguments have been fully considered and are persuasive. The rejection has been withdrawn.
35 USC 103
Applicant’s arguments with respect to 35 USC 103 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 § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1, 3-11, 14, 16-19, 21, and 24 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Tashev et al. (“Data Driven Suppression Rule for Speech Enhancement”).
Regarding claim 1, Tashev discloses: 1. An apparatus, comprising: at least one processor; and
at least one non-transitory memory storing instructions that, when executed with the at least one processor, cause the apparatus at least to: ("This converts the design problem into modeling … which we can tackle with large-data machine-learning techniques." Pg. 3, para 5 – machine learning implies a processor and memory)
use a machine learning program to obtain two or more outputs (“p = [p1, p2, …, p9]” pg. 3, equation (5) – the 9 parameters are obtained by a machine learning process from the training data – see pg. 4, section III., especially “The design problem then becomes one of finding the values of the parameters vector p that give a model with the closest fit to the desired suppression gain over all the training data.” and pg. 4, section A.)
for at least one of a plurality of different frequency bands, ("Taking the discrete Fourier transform on N windowed intervals of length 2K yields K frequency bins per frame: Yk = Xk + Dk, where all these quantities are complex. Noise reduction may be viewed as the application of a suppression rule, or nonnegative real valued gain Hk, to each bin k of the observed signal spectrum Yk, in order to form an estimate ˆ k X of the original signal spectrum: ˆ k kk X HY . " pg. 2, section A. The Data)
wherein respective outputs of the two or more outputs prioritize different output objectives; (pg. 3, equation (5) shows that p2, p5, and p8 prioritize the a priori SNR and p3, p6, and p9 prioritize the a posteriori SNR. See also “This spectral magnitude estimator provides noise suppression while maintaining lower distortions and fewer artifacts.” Pg. 2, col 2, para 4; see also Fig. 1)
obtain one or more tuning parameters, wherein at least one of the one or more tuning parameters is configured to balance between the two or more outputs; and ("By defining the problem in such a way we are able to tailor the suppression rule for specific set of input data (SNRs, type of noise), stressing more or less different regions of the enhancement space. In all of these cases we find a suppression rule optimal for the specific problem. We have the flexibility to include additional components in the optimization criterion, such as MSE and/or log-MSE. The full optimization in this case looks like:
PNG
media_image1.png
86
684
media_image1.png
Greyscale
With this additional flexibility we can change the weights of the different components of the optimization criterion. Theoretically w = [0,1, 0] should be equivalent to the ST MSE suppression rule, and w = [0,0,1] should be equivalent to the ST log-MSE suppression rule. In practice the MSE error (6) has a large effect where the speech signal has higher amplitudes, while log-MSE (7) gives more weight to the regions with lower amplitudes. This allows fine tuning of the received result, still keeping PESQ as the component with the highest weight. " pg. 3, col 2, para 3-4. Weight vector w contains 3 tuning parameters to balance between objectives, and each parameter affects the outputs.)
process the two or more outputs to determine at least one uncertainty value ("The equations for P, Q, and R include linear terms of the prior and posterior probabilities to allow simple rotations/skews." Pg. 3 para 7 – equation (5) shows that P, Q, and R, include probabilities (which are simply the inverse of uncertainty). P, Q, and R are based on p.)
and a gain coefficient for the at least one of the plurality of different frequency bands, (pg. 3, equation (5), “H” is a gain coefficient. )
wherein the at least one uncertainty value provides a measure of uncertainty for the gain coefficient; (pg. 3, equation (5) shows that the values P, Q, and R determine H (gain). )
and adjust the gain coefficient based, at least partially, on the at least one uncertainty value (pg. 3, equation (5) shows that H is based at least partially on P, Q, and R )
and the one or more tuning parameters; (pg. 3, equation shows that P, Q, and R are based on parameters p which are based on tuning parameters w (equation (9)) )
wherein the adjusted gain coefficient is configured to be applied to a signal associated with at least one microphone output signal within the at least one of the plurality of different frequency bands to control noise suppression for speech audibility. (pg. 4, Table I shows experimental results of applying the gain for noise suppression for speech audibility. )
Regarding claim 3, Tashev discloses: 3. An apparatus as claimed in claim 1, wherein the two or more outputs of the machine learning program comprise gain coefficients that correspond to the different output objectives. (As mapped above, H is a gain coefficient that corresponds to the objectives of suppressing noise while maintaining lower distortions and fewer artifacts (pg. 2 section E). Further, there is a separate coefficient for each frequency bin (pg. 2 section A) which can be read as a further output objective of suppressing noise and maintaining lower distortion in that frequency range. )
Regarding claim 4, Tashev discloses: 4. An apparatus as claimed in claim 1, wherein the instructions, when executed with the at least one processor, cause the apparatus to adjust noise reduction relative to speech distortion. (" This spectral magnitude estimator provides noise suppression while maintaining lower distortions and fewer artifacts. The shape of this suppression rule is shown in Figure 1." Pg. 2 section E )
Regarding claim 5, Tashev discloses: 5. An apparatus as claimed in claim 1, wherein the signal comprises at least one of: speech; or noise. (" Each data file contains ten randomly selected utterances from different speakers. To the clean speech files we added stationary Hoth noise…" pg. 4, section A )
Regarding claim 6, Tashev discloses: 6. An apparatus as claimed in claim 1, wherein the instructions, when executed with the at least one processor, cause the apparatus to: configure different functions corresponding to the different output objectives for the machine learning program, wherein the different functions comprise different values for one or more objective weight parameters. ("By defining the problem in such a way we are able to tailor the suppression rule for specific set of input data (SNRs, type of noise), stressing more or less different regions of the enhancement space. In all of these cases we find a suppression rule optimal for the specific problem. We have the flexibility to include additional components in the optimization criterion, such as MSE and/or log-MSE. The full optimization in this case looks like:
PNG
media_image1.png
86
684
media_image1.png
Greyscale
With this additional flexibility we can change the weights of the different components of the optimization criterion. Theoretically w = [0,1, 0] should be equivalent to the ST MSE suppression rule, and w = [0,0,1] should be equivalent to the ST log-MSE suppression rule. In practice the MSE error (6) has a large effect where the speech signal has higher amplitudes, while log-MSE (7) gives more weight to the regions with lower amplitudes. This allows fine tuning of the received result, still keeping PESQ as the component with the highest weight. " pg. 3, col 2, para 3-4. Weight vector w contains 3 tuning parameters to balance between objectives, and each parameter affects the outputs. The weights determine the amount of suppression in different areas while giving a separate weight to speech quality.)
Regarding claim 7, Tashev discloses: 7. An apparatus as claimed in claim 6,wherein the instructions, when executed with the at least one processor, cause the apparatus to: cause a first value for the one or more objective weight parameters to prioritize noise reduction over avoiding speech distortion; and cause a second value for the one or more objective weight parameters to prioritize avoiding the speech distortion over the noise reduction. (
PNG
media_image1.png
86
684
media_image1.png
Greyscale
pg. 3, equation 9 shows that w1 prioritizes avoiding speech distortion while w2 and w3 prioritize noise reduction.)
Regarding claim 8, Tashev discloses: 8. An apparatus as claimed in claim 1,wherein the instructions, when executed with the at least one processor, cause the apparatus to: determine the gain coefficient based on a mean of the two or more outputs of the machine learning program and the at least one uncertainty value. (“For each suppression rule, we found the model parameters, p, that minimize the mean-squared difference in the log domain between the desired (classic) suppression rule and the parameterized suppression model (Eq. (5):" pg. 4 section B. )
Regarding claim 9, Tashev discloses: 9. An apparatus as claimed in claim 1,wherein the at least one uncertainty value is based on a difference between the two or more outputs of the machine learning program. (“For each suppression rule, we found the model parameters, p, that minimize the mean-squared difference in the log domain between the desired (classic) suppression rule and the parameterized suppression model (Eq. (5):" pg. 4 section B. )
Regarding claim 10, Tashev discloses: 10. An apparatus as claimed in claim 1,wherein the instructions, when executed with the at least one processor, cause the apparatus to: cause the one or more tuning parameters to control one or more variables of the adjustment of the gain coefficient. (pg. 3, equations 5 and 9, show that w controls variables that adjust the gain coefficient H.)
Regarding claim 11, Tashev discloses: 11. An apparatus as claimed in claim 1, wherein the adjustment of the gain coefficient with the at least one uncertainty value and the one or more tuning parameters comprises a weighting of the two or more outputs of the machine learning program. (pg. 3, equation 9, shows that w weights the outputs p.)
Regarding claim 14, Tashev discloses: 14. An apparatus as claimed in claim 1,wherein the instructions, when executed with the at least one processor, cause the apparatus to: cause the machine learning program to receive a plurality of inputs, for one or more of the plurality of different frequency bands, wherein the plurality of inputs comprise one or more of: an acoustic echo cancellation signal; a loudspeaker signal; a microphone signal; or a residual error signal. (" Each data file contains ten randomly selected utterances from different speakers. To the clean speech files we added stationary Hoth noise…" pg. 4, section A – speech and noise are a plurality of inputs, and the speech comprises at least a microphone signal.)
Regarding claim 16, Tashev discloses: 16. An apparatus as claimed in claim 1, wherein the instructions, when executed with the at least one processor, cause the apparatus to: adjust the one or more tuning parameters based on one or more of: a user input; a determined use case; a determined change in echo path; determined acoustic echo cancellation measurements; wind estimates; signal noise ratio estimates; spatial audio parameters; voice activity detection; nonlinearity estimation; or clock drift estimations. ("The final row of Table 1 (“Optimal quality”) contains the result optimized using Eq. (9). In this particular case we used weight vector w > @ 1.0,1.0, 0.01 which, considering the values range of each parameter, gives most of the weight to PESQ.” Pg. 5 section C – this is a user input)
Regarding claim 17, Tashev discloses: 17. An apparatus as claimed in claim 1, wherein the instructions, when executed with the at least one processor, cause the apparatus to use the machine learning program to obtain two or more outputs for the plurality of different frequency bands. ("Noise reduction may be viewed as the application of a suppression rule, or nonnegative real-valued gain Hk, to each bin k of the observed signal spectrum Yk, in order to form an estimate ˆ k X of the original signal spectrum: ˆ k kk X HY ." pg. 2 section A )
Regarding claim 18, Tashev discloses: 18. An apparatus as claimed in claim 1,wherein the signal associated with at least one microphone output signal comprises at least one of: a raw at least one microphone output signal; a processed at least one of microphone output signal; or a residual error signal. (" Each data file contains ten randomly selected utterances from different speakers. To the clean speech files we added stationary Hoth noise…" pg. 4, section A – the clean speech is a raw output, and adding noise to the clean speech is a processing of the speech.)
Regarding claim 19, Tashev discloses: 19. An apparatus as claimed in claim 1,wherein the signal associated with at least one microphone output signal is a frequency domain signal. ("The process of conversion to the frequency domain and back was performed using the scripts provided in Tashev’s book [15].” Pg. 4, section A )
Claim 21 is a method claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale.
Claim 24 is a device claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 12-13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tashev et al. in view of Dickins et al. (US 20140126745 A1).
Regarding claim 12, Tashev does not disclose explicitly that different tuning parameters are used for different frequency bands, although it is clear that the model of Tashev would support this as each frequency bin has its own gain Hk. (“Noise reduction may be viewed as the application of a suppression rule, or nonnegative realvalued gain Hk, to each bin k of the observed signal spectrum Yk, in order to form an estimate ˆ k X of the original signal spectrum: ˆ k kk X HY .” Pg. 2 section A)
Dickins discloses: 12. An apparatus as claimed in claim 1,wherein the instructions, when executed with the at least one processor, cause the apparatus to: use different tuning parameters for respective ones of the plurality of different frequency bands. ("[0228] In a further embodiment, a band-dependent weighting factor can be introduced into the echo update voice-activity detector 125 such that the individual band contributions based on the instantaneous signal to noise ratio are weighted across frequency for their contribution to the detection of signal activity...."
Tashev and Dickins are considered analogous art to the claimed invention because they disclose noise reduction systems. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tashev with band-dependent weights as taught by Dickins. This would have been beneficial for specific applications or to enhance sensitivity to certain expected stimulus. (Dickins [0228])
Regarding claim 13, Tashev does not disclose the additional limitations.
Dickins discloses: 13. An apparatus as claimed in claim 1,wherein the instructions, when executed with the at least one processor, cause the apparatus to: use different tuning parameters for different time intervals. ("[0143] In some embodiments, the beamformer 107 (and beamforming step 207) can include adaptive tracking of the spatial selectivity over time, in which case the beamformer gains (also called beamformer weights) are updated as appropriate to track some spatial selectivity in the estimated position of the source of interest…" )
Tashev and Dickins are considered analogous art to the claimed invention because they disclose noise reduction systems. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tashev with time-dependent weights as taught by Dickins. This would have been beneficial if the noise is varying over time. (Dickins [0208])
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tashev et al. in view of De Vries et al. (US 20040047474 A1).
Regarding claim 15, Tashev does not disclose the additional limitations.
De Vries discloses: 15. An apparatus as claimed in claim 1, wherein the machine learning program comprises a neural network circuit. ("[0068] The above-mentioned adaptation scheme is a well-known "machine learning" type of application. We choose an environmental classifier that controls the parameters of the noise suppression agent or agents 5, 10, 15 and 20 such that the target y(t)=s(t)+g*n(t) is obtained as closely as possible for the inputs x(t)=s(t)+n(t). The classifier 25 is therefore a parameterized learning machine such as a Hidden Markov Model, neural network, fuzzy logic machine or any other machine with adaptive parameters and can be trained by learning mechanisms that are well-known in the art such as back propagation..." )
Tashev and De Vries are considered analogous art to the claimed invention because they disclose noise reduction systems. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tashev with a machine learning program for performing calculations. This combination falls under simple substitution of one known element for another to obtain predictable results or use of known technique to improve similar devices (methods, or products) in the same way. See MPEP 2141, KSR, 550 U.S. at 418, 82 USPQ2d at 1396.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JON C MEIS whose telephone number is (703)756-1566. The examiner can normally be reached Monday - Thursday, 8:30 am - 5:30 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Hai Phan can be reached on 571-272-6338. 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.
/JON CHRISTOPHER MEIS/Examiner, Art Unit 2654
/HAI PHAN/Supervisory Patent Examiner, Art Unit 2654