DETAILED ACTION
Notice of 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 .
Continued Examination Under 37 CFR 1. 114
2. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1 .114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 07/17/2026 has been entered.
Response to Amendment
3. Amendment filed 07/17/2026 has been considered by Examiner. Claims 1, 10 and 18 have been amended. Claims 1-20 are pending, and likewise Claims 1- 20 have been examined.
Response to Arguments
Applicant’s amendments and arguments filed 07/17/2026, with respect to claim(s) 1-20 have been fully considered, with respect to claim(s) 1-20, 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 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 of this title, 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.
Claims 1, 2, 8 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, in view of Su et al. ( US 6122611), hereinafter referenced as Su, further in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath.
Regarding Claim 1, Chang teaches a method comprising: receiving an indication of a lost audio packet at a receive side of a neural network audio codec system that includes an audio encoder and an audio decoder, [wherein the lost audio packet comprises an index of a codeword] that is representative of a portion of speech audio presented to the audio encoder (Chang: Para.[0052],[0053], Fig. 4 illustrates a convolutional neural network consisting of encoder and decoder. A target frame 420 of a voice signal, may indicate a frame including a packet in which a loss has occurred) ;
predicting the index of the codeword in the lost audio packet to obtain a predicted index [ wherein the predicted index is predicted by conditioning, recursively, on previously predicted indices of codewords from prior lost audio packets] ( Chang: Para.[0062], Fig. 6, when a packet loss occurs in a voice signal encoded by a codec, lost packet information estimated based on a trained generative model);
deriving a predicted embedding vector from the predicted index ( Chang: Para.[0062], Fig. 6, a feature vector corresponding to the lost packet (i. e., lost frame) may be estimated (or generated) based on a feature vector (i.e., an FFT coefficient and a phase) extracted from P frames prior to the loss);
Chang while teaching the method of claim 1, fails to explicitly teach the claimed, wherein the lost audio packet comprises an index of a codeword, [predicting the index of the codeword in the lost audio packet to obtain a predicted index ], wherein the predicted index is predicted by conditioning, recursively, on previously predicted indices of codewords from prior lost audio packets; and decoding, by the audio decoder, the predicted embedding vector to generate an audio output.
However, Su does teach the claimed, audio packet comprises an index of a codeword ( Su: Column 2, lines 4-19, column 5, lines 1-8, Fig. 7, decoder circuit 502 receives speech signal 414 which includes index of a codeword);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Su’s teaching of a system and method to improve the quality of coded speech coexisting with background noise , into the system and method of a packet loss concealment method and apparatus using a generative adversarial network, taught by Chang, because, the improved speech signal would sound natural and realistic to the human ear to enhance the experience.(Su, Column 3, lines 1-25).
Chang in view of Su, while teaching the method of claim 1, fail to explicitly teach the claimed, predicting the index of the codeword in the lost audio packet to obtain a predicted index , wherein the predicted index is predicted by conditioning, recursively, on previously predicted indices of codewords from prior lost audio packets; and decoding, by the audio decoder, the predicted embedding vector to generate an audio output.
However, Yahampath does teach the claimed, predicting the index of the codeword in the lost audio packet to obtain a predicted index , wherein the predicted index is predicted by conditioning, recursively, on previously predicted indices of codewords from prior lost audio packets ( Yahampath: Section II, Para.[1], Section III, Para.[2], Section III.D, Fig. 1, an algorithm for designing linear prediction-based two-channel multiple-description predictive-vector quantizers (MD-PVQs) for packet-loss channels is presented. MDVQ encoder generates two quantization indices ( codewords) by quantizing the prediction error. The knowledge of the loss probabilities and the channel loss patterns are used. Once, each system component has been optimized for its respective input training set, these training sets has been recomputed for the next design iteration using the updated system. The procedure is repeated until the change in the average distortion between two consecutive iterations is negligible. Since, in each design iteration, encoder, codebook, decoder and predictor is optimized to their respective inputs, the prediction error and the quantization error generally decrease with each iteration);
and decoding, by the audio decoder, the predicted embedding vector to generate an audio output (Yahampath: Section II. Para.[2], page 750, Fig. 1, The MD-PVQ decoder computes its output a X’n=Un’+ x’n, where x’n is the prediction at the decoder)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Yahampath’s teaching of designing linear prediction-based two-channel multiple-description predictive-vector quantizers (MD-PVQs) for packet-loss channels, into the system and method of a packet loss concealment method and apparatus using a generative adversarial network, taught by Chang in view of Su, because, this algorithm iteratively improves the encoder partition, the set of multiple description codebooks, and the linear predictor for a given channel loss probability. (Yahampath, Abstract, Section I).
Regarding Claim 2, Chang in view of Su, further in view of Yahampath teach the method of claim 1. Chang further teaches, wherein predicting the index of the codeword of the lost audio packet is based on a history of previously received indices of codewords ( Chang: Para.[0046], The generative model may restore the lost packet by estimating packet information corresponding to a lost frame based on feature vectors extracted from a predetermined number of previous frames ( history)).
Regarding Claim 8, Chang in view of Su, further in view of Yahampath teach the method of claim 1. Chang further teaches, wherein predicting the index of the codeword of the lost audio packet is based on successfully received indices of codewords ( Chang: Para.[0046], The generative model may restore the lost packet by estimating packet information corresponding to a lost frame based on feature vectors extracted from a predetermined number of previous frames).
Regarding Claim 9, Chang in view of Su, further in view of Yahampath teach the method of claim 8. Chang further teaches, further comprising predicting the index of the codeword of the lost audio packet in a recursive manner (Chang: Para.[0060],[0067], Fig. 6 illustrates a method for packet loss concealment with a neural network, which is trained in a deep learning-based recursive model to improve performance).
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, in view of Su et al. ( US 6122611), hereinafter referenced as Su, further in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Cheluvaraja et al. (US 20150255075 A1), hereinafter referenced as Cheluvaraja.
Regarding Claim 3, Chang in view of Su, further in view of Yahampath teach the method of claim 2. Chang in view of Su, further in view of Yahampath fail to explicitly teach the claimed, further comprising predicting the index of the codeword of the lost audio packet by maximizing a probability, in a previously-trained machine learning process, that the predicted index is a likely next index given the history of previously received indices of codewords.
However, Cheluvaraja does teach the claimed, further comprising predicting the index of the codeword of the lost audio packet by maximizing a probability, in a previously-trained machine learning process, that the predicted index is a likely next index given the history of previously received indices of codewords ( Cheluvaraja: Para.[0035]-[0039],Fig. 2, the most likely feature values of the succeeding packet are found by maximizing a probability function where a previously trained acoustic model may be utilized in the prediction. The best estimate for x(t+1) is given by maximizing P(x(t+1)|x(1), x(2) . . . x(t)) over all possible x(t+1)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Cheluvaraja’s teaching of system and method for the correction of packet loss in audio in automatic speech recognition system, into the system and method, taught by Chang in view of Su, further in view of Yahampath, because, this would improve in overall ASR recognition accuracy.( Cheluvaraja [ Para.[0029]).
Claims 4, 5 are rejected under 35 U.S.C. 103 as being unpatentable over Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, in view of Su et al. ( US 6122611), hereinafter referenced as Su, further in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Cheluvaraja et al. (US 20150255075 A1), hereinafter referenced as Cheluvaraja, further in view of Park et al. (US 20230274731 A1), hereinafter referenced as Park.
Regarding Claim 4, Chang in view of Su, further in view of Yahampath, further in view of Cheluvaraja teach the method of claim 3. Chang in view of Su, further in view of Yahampath, further in view of Cheluvaraja fail to explicitly teach the claimed, wherein the previously-trained machine learning process was trained based on classification losses.
However, Park does teach the claimed, wherein the previously-trained machine learning process was trained based on classification losses ( Park: Para.[0029], [0064], during the training, the neural network analyzes the training sample and then generates an output or prediction which is compared to the predefined target output (i.e., the label) to determine a loss using a loss function, which can be regression loss, hinge loss, multi-class loss, etc.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Park’s teaching of a method for training a neural network , into the system and method, taught by Chang in view of Su, further in view Yahampath and Cheluvaraja, because, this would optimize the training of the neural network.( Park [ Para.[0011], [0048]).
Regarding Claim 5, Chang in view of Su, further in view of Yahampath, further in view of Cheluvaraja, further in view of Park teach the method of claim 4. Park further teaches, wherein the classification losses comprise cross entropy evaluation ( Park: Para.[0029], [0030], during the training, the neural network analyzes the training sample and then generates an output or prediction which is compared to the predefined target output (i.e., the label) to determine a loss using a loss function, such as cross entropy based loss)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Park’s teaching of a method for training a neural network , into the system and method, taught by Chang in view of Su, in view of Yahampath and Cheluvaraja, because, this would optimize the training of the neural network.( Park [ Para.[0011], [0048]).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, in view of Su et al. ( US 6122611), hereinafter referenced as Su, further in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Cheluvaraja et al. (US 20150255075 A1), hereinafter referenced as Cheluvaraja, further in view of Kuroiwa et al. (JP 2004272128 A), hereinafter referenced as Kuroiwa.
Regarding Claim 7, Chang in view of Su, further in view of Yahampath, further in view of Cheluvaraja teach the method of claim 3. Chang in view of Su, further in view of Yahampath, further in view of Cheluvaraja fail to explicitly teach the claimed, wherein the previously-trained machine learning process is a language model that is trained to predict codeword indices to codewords in a codebook.
However, Kuroiwa does teach the claimed, wherein the previously-trained machine learning process is a language model that is trained to predict codeword indices to codewords in a codebook ( Kuroiwa: Para.[0038], the period 102 during which there is a packet loss is complemented by the outputs X .sub.4 to X .sub.6 of the voice synthesizer 52. This complementation is performed using not only an acoustic model but also a language model called a language model. For example, even if a voice signal for one syllable is completely lost, the syllable can be statistically predicted from contexts before and after. Then, speech synthesis is performed using the predicted syllables to compensate for the missing portion of the speech. Para.[0045], In the audio signal restoration device 30 according to the above-described embodiment, a natural language is assumed as a language model. However, the language model is not limited to such a natural language. For example, each code in a codebook used in a codec or the like can be regarded as one phoneme, and the language model 44 can be created from statistical information for the phonemes. Even when a language model of a natural language is used, a language model may be created by treating a section having a statistical unit as a word, instead of using a so-called word as a unit].
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Kuroiwa’s teaching of a restoring device for voice signal, into the system and method, taught by Chang in view of Su, further in view of Yahampath and Cheluvaraja, because, this would eliminate voice interruption caused by packet loss .( Kuroiwa [ Para.[0005]).
Claims 10,16, 17 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Su et al. ( US 6122611 A), hereinafter referenced as Su, in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang.
Regarding Claim 10, Su teaches a method comprising: receiving a sequence of audio packets representing speech audio, each audio packet including an index of a codeword that is representative of a portion of the speech audio ( Su: Column 2, lines 4-19, column 5, lines 1-8, Fig. 7, decoder circuit 502 receives speech signal 414 which includes index of a codeword);
predicting an index of a codeword for a current audio packet based on one or more previous audio packets in the sequence of audio packets to obtain a predicted index ( Su: Column 6, lines 16-22, 36-40, Fig. 7, linear prediction coefficient synthesis filter circuit 706 receives the encoded linear prediction coefficients, contained within coded speech signal 414);
Su, while teaching the method of claim 10, fails to explicitly teach the claimed, wherein the predicted index is predicted by conditioning, recursively, on previously predicted indices of codewords from prior lost audio packets; deriving a predicted embedding vector from the predicted index; and decoding the predicted embedding vector to generate audio output.
However, Yahampath does teach the claimed, wherein the predicted index is predicted by conditioning, recursively, on previously predicted indices of codewords from prior lost audio packets ( Yahampath: Section II, Para.[1], Section III, Para.[2], Section III.D, Fig. 1, an algorithm for designing linear prediction-based two-channel multiple-description predictive-vector quantizers (MD-PVQs) for packet-loss channels is presented. MDVQ encoder generates two quantization indices ( codewords) by quantizing the prediction error. The knowledge of the loss probabilities and the channel loss patterns are used. Once, each system component has been optimized for its respective input training set, these training sets has been recomputed for the next design iteration using the updated system. The procedure is repeated until the change in the average distortion between two consecutive iterations is negligible. Since, in each design iteration encoder, codebook, decoder and predictor is optimized to their respective inputs, the prediction error and the quantization error generally decrease with each iteration);
and decoding the predicted embedding vector to generate audio output (Yahampath: Section II. Para.[2], page 750, Fig. 1, The MD-PVQ decoder computes its output a X’n=Un’+ x’n, where x’n is the prediction at the decoder)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Yahampath’s teaching of designing linear prediction-based two-channel multiple-description predictive-vector quantizers (MD-PVQs) for packet-loss channels, into the system and method to improve the quality of coded speech coexisting with background noise, taught by Su, because, this algorithm iteratively improves the encoder partition, the set of multiple description codebooks, and the linear predictor for a given channel loss probability. (Yahampath, Abstract, Section I).
Su in view of Yahampath, while teaching the method of claim 10, fails to explicitly teach the claimed, deriving a predicted embedding vector from the predicted index;
However, Chang does teach the claimed, deriving a predicted embedding vector from the predicted index ( Chang: Para.[0063], Fig. 6, at 658, a feature vector may be estimated (or generated) from the signal );
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chang’s teaching of a packet loss concealment method and apparatus using a generative adversarial network , into the system and method, taught by Su in view of Yahampath, because, this would enhance the sound quality of the voice signal.(Chang [ Para.[0005]-[0012]).
Regarding Claim 16, Su in view of Yahampath, further in view of Chang teach the method of claim 10. Chang further teaches, wherein predicting the index of the codeword is based on successfully received indices of codewords ( Chang: Para.[0046], The generative model may restore the lost packet by estimating packet information corresponding to a lost frame based on feature vectors extracted from a predetermined number of previous frames).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chang’s teaching of a packet loss concealment method and apparatus using a generative adversarial network , into the system and method, taught by Su in view of Yahampath, because, this would enhance the sound quality of the voice signal.(Chang [ Para.[0005]-[0012]).
Regarding Claim 17, Su in view of Yahampath, further in view of Chang teach the method of claim 16. Chang further teaches, further comprising predicting the index of the codeword in a recursive manner (Chang: Para.[0060],[0067], Fig. 6 illustrates a method for packet loss concealment with a neural network, which is trained in a deep learning-based recursive model to improve performance).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chang’s teaching of a packet loss concealment method and apparatus using a generative adversarial network , into the system and method, taught by Su in view of Yahampath, because, this would enhance the sound quality of the voice signal.(Chang [ Para.[0005]-[0012]).
Regarding Claim 18, Su teaches a device comprising: an interface configured to enable network communications ( Su: Column 4, lines 23-26, Fig. 4, communication network 406);
a memory; and one or more processors coupled to the interface and the memory ( Su: Column 4, lines 47-54, Fig. 4,storage device 404 are a random access memory (RAM) unit, a floppy disk, a hard drive memory unit, connected to analysis unit 402 ( as processor)),
and configured to: receive a sequence of audio packets representing speech audio, each audio packet including an index of a codeword that is representative of a portion of the speech audio( Su: Column 2, lines 4-19, column 5, lines 1-8, Fig. 7, decoder circuit 502 receives speech signal 414 which includes index of a codeword);
predict an index of a codeword for a current audio packet based on one or more previous audio packets in the sequence of audio packets to obtain a predicted index ( Su: Column 6, lines 16-22, 36-40, Fig. 7, linear prediction coefficient synthesis filter circuit 706 receives the encoded linear prediction coefficients, contained within coded speech signal 414);
Su, while teaching the device of claim 18, fails to explicitly teach the claimed, wherein the predicted index is predicted by conditioning , recursively, on previously predicted indices of codewords from prior lost audio packets; deriving a predicted embedding vector from the predicted index; and decoding the predicted embedding vector to generate audio output.
However, Yahampath does teach the claimed, wherein the predicted index is predicted by conditioning , recursively, on previously predicted indices of codewords from prior lost audio packets ( Yahampath: Section II, Para.[1], Section III, Para.[2], Section III.D, Fig. 1, an algorithm for designing linear prediction-based two-channel multiple-description predictive-vector quantizers (MD-PVQs) for packet-loss channels is presented. MDVQ encoder generates two quantization indices ( codewords) by quantizing the prediction error. The knowledge of the loss probabilities and the channel loss patterns are used. Once, each system component has been optimized for its respective input training set, these training sets has been recomputed for the next design iteration using the updated system. The procedure is repeated until the change in the average distortion between two consecutive iterations is negligible. Since, in each design iteration encoder, codebook, decoder and predictor is optimized to their respective inputs, the prediction error and the quantization error generally decrease with each iteration);
and decoding the predicted embedding vector to generate audio output (Yahampath: Section II. Para.[2], page 750, Fig. 1, The MD-PVQ decoder computes its output a X’n=Un’+ x’n, where x’n is the prediction at the decoder)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Yahampath’s teaching of designing linear prediction-based two-channel multiple-description predictive-vector quantizers (MD-PVQs) for packet-loss channels, into the system and method to improve the quality of coded speech coexisting with background noise, taught by Su, because, this algorithm iteratively improves the encoder partition, the set of multiple description codebooks, and the linear predictor for a given channel loss probability. (Yahampath, Abstract, Section I).
Su in view of Yahampath, while teaching the device of claim 18, fails to explicitly teach the claimed, deriving a predicted embedding vector from the predicted index;
However, Chang does teach the claimed, deriving a predicted embedding vector from the predicted index ( Chang: Para.[0063], Fig. 6, at 658, a feature vector may be estimated (or generated) from the signal );
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chang’s teaching of a packet loss concealment method and apparatus using a generative adversarial network , into the system and method, taught by Su in view of Yahampath, because, this would enhance the sound quality of the voice signal.(Chang [ Para.[0005]-[0012]).
Claims 11,19 are rejected under 35 U.S.C. 103 as being unpatentable over Su et al. ( US 6122611), hereinafter referenced as Su, in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, further in view of Cheluvaraja et al. (US 20150255075 A1), hereinafter referenced as Cheluvaraja.
Regarding Claim 11, Su in view of Yahampath, further in view of Chang teach the method of claim 10. Su in view of Yahampath, further in view of Chang fail to explicitly teach the claimed, further comprising predicting the index of the codeword by maximizing a probability, in a previously-trained machine learning process, that a predicted index of the codeword is a likely next index given the history of previously received indices of codewords.
However, Cheluvaraja does teach the claimed, further comprising predicting the index of the codeword by maximizing a probability, in a previously-trained machine learning process, that a predicted index of the codeword is a likely next index given the history of previously received indices of codewords ( Cheluvaraja: Para.[0035]-[0039],Fig. 2, the most likely feature values of the succeeding packet are found by maximizing a probability function where a previously trained acoustic model may be utilized in the prediction. The best estimate for x(t+1) is given by maximizing P(x(t+1)|x(1), x(2) . . . x(t)) over all possible x(t+1)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Cheluvaraja’s teaching of system and method for the correction of packet loss in audio in automatic speech recognition system, into the system and method, taught by Su in view of Yahampath, further in view of Chang, because, this would improve in overall ASR recognition accuracy.( Cheluvaraja [ Para.[0029]).
Claim 19 is device claim performing the steps in method claim 11 above and as such, claim 19 is similar in scope and content to claim 11 and therefore, claim 19 is rejected under similar rationale as presented against claim 11 above.
Claims 12, 13 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Su et al. ( US 6122611), hereinafter referenced as Su, in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, further in view of Cheluvaraja et al. (US 20150255075 A1), hereinafter referenced as Cheluvaraja, further in view of Park et al. (US 20230274731 A1), hereinafter referenced as Park.
Regarding Claim 12, Su in view of Yahampath, further in view of Chang, further in view of Cheluvaraja teach the method of claim 11. Su in view of Yahampath, further in view of Chang, further in view of Cheluvaraja fail to explicitly teach the claimed, wherein the previously-trained machine learning process was trained based on classification losses.
However, Park does teach the claimed, wherein the previously-trained machine learning process was trained based on classification losses ( Park: Para.[0029], [0064], during the training, the neural network analyzes the training sample and then generates an output or prediction which is compared to the predefined target output (i.e., the label) to determine a loss using a loss function, which can be regression loss, hinge loss, multi-class loss, etc.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Park’s teaching of a method for training a neural network , into the system and method, taught by Su in view of Yahampath, further in view of Chang, further in view of Cheluvaraja, because, this would optimize the training of the neural network.( Park [ Para.[0011], [0048]).
Claim 20 is device claim performing the steps in method claim 12 above and as such, claim 20 is similar in scope and content to claim 12 and therefore, claim 20 is rejected under similar rationale as presented against claim 12 above.
Regarding Claim 13, Su in view of Yahampath, further in view of Chang, further in view of Cheluvaraja and Park teach the method of claim 11. Park further teaches, wherein the classification losses comprise cross entropy evaluation ( Park: Para.[0029], [0030], during the training, the neural network analyzes the training sample and then generates an output or prediction which is compared to the predefined target output (i.e., the label) to determine a loss using a loss function, such as cross entropy based loss)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Park’s teaching of a method for training a neural network , into the system and method, taught by Su in view of Yahampath, further in view of Chang and Cheluvaraja, because, this would optimize the training of the neural network.( Park [ Para.[0011], [0048]).
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Su et al. ( US 6122611), hereinafter referenced as Su, in view of Yahampath et al. ( Multiple-Description Predictive-Vector Quantization With Applications to Low Bit-Rate Speech Coding Over Networks, IEEE, 2007), hereinafter referenced as Yahampath, further in view of Chang et al. ( US 20190051310 A1), hereinafter referenced as Chang, further in view of Cheluvaraja et al. (US 20150255075 A1), hereinafter referenced as Cheluvaraja, further in view of Kuroiwa et al. (JP 2004272128 A), hereinafter referenced as Kuroiwa.
Regarding Claim 15, Su in view of Yahampath, further in view of Chang, further in view of Cheluvaraja teach the method of claim 11. Su in view of Yahampath, further in view of Chang, further in view of Cheluvaraja fail to explicitly teach the claimed, wherein the previously-trained machine learning process is a language model that is trained to predict codeword indices to codewords in a codebook.
However, Kuroiwa does teach the claimed, wherein the previously-trained machine learning process is a language model that is trained to predict codeword indices to codewords in a codebook ( Kuroiwa: Para.[0038], the period 102 during which there is a packet loss is complemented by the outputs X .sub.4 to X .sub.6 of the voice synthesizer 52. This complementation is performed using not only an acoustic model but also a language model called a language model. For example, even if a voice signal for one syllable is completely lost, the syllable can be statistically predicted from contexts before and after. Then, speech synthesis is performed using the predicted syllables to compensate for the missing portion of the speech. Para.[0045], In the audio signal restoration device 30 according to the above-described embodiment, a natural language is assumed as a language model. However, the language model is not limited to such a natural language. For example, each code in a codebook used in a codec or the like can be regarded as one phoneme, and the language model 44 can be created from statistical information for the phonemes. Even when a language model of a natural language is used, a language model may be created by treating a section having a statistical unit as a word, instead of using a so-called word as a unit].
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Kuroiwa’s teaching of a restoring device for voice signal, into the system and method, taught by Su, in view of Yahampath, further in view of Chang and Cheluvaraja, because, this would eliminate voice interruption caused by packet loss .( Kuroiwa [ Para.[0005]).
Allowable Subject Matter
Claims 6 and 14 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NADIRA SULTANA whose telephone number is (571)272-4048. The examiner can normally be reached M-F,7:30 am-5:00pm.
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 on (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.
/NADIRA SULTANA/Examiner, Art Unit 2653
Prosecution Insights