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 .
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.
Claims 1, 4, 6, 10,15 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Bhattacharya et al. (US 2022/0375462A1), hereinafter “ Bhattacharya’.
As to Claim 1, Bhattacharya teaches a)(computer-implemented method for processing audio data in an audio device by using a neural network, ( a method and apparatus for conditioning neural networks that dynamically modifies the manner in which non-linear activations process inputs into outputs based on conditioning vectors, [0007], [0060]-[0063], Figure 2) the neural network is defined by its topology including its number of layers and its number of nodes ( implicit, as neural network is a machine learning model consisting of interconnected nodes or neurons and multiple layers) where each node has an activation function ( implicit to the neural network model consisting of nodes , each node governed by a threshold and an activation function), said method comprising: obtaining first audio data,( S200, Figure 10, [0090]) based on an input, adapting the activation function(s) of the one or more nodes of the neural network,( S100, S102, Figure 4 and “ input b” on Figure 2, [0081] and Figure 10, [0090] “ applying conditioning vector”) while maintaining the topology of the neural network,( does not significantly increase the model parameters.[0065) thereby allowing the neural network to adapt in a computationally efficient way,( the conditioning vectors are used to dynamically modify the activation functions, [0065], S104, Figure 4, Figure 10 [0090] generating a conditioned activation function for the at least one layer (S206); ) processing the first audio data, into processed audio data, by using the neural network with the adapted activation function(s)( S108, Figure 4), [0090], Figure 10, and outputting the processed audio data [0088] and [0090] on Figure 10, S208]).
As to Claim 4, Bhattacharya teaches the limitations of Claim 1 and wherein the activation function code bank comprises at least one or more of the following activation functions: ReLU; tanh; sigmoid; linear; and/or identity, [0082] teaches the initial activation function may be non-linear functions to be applied to a linear layer of the network. Of course, it will be appreciated that the initial activation functions may include linear functions also. By way of example, at least one layer may be a Dense Linear layer, and the identified initial activation functions may be non-linear functions {elu, exponential, hard sigmoid, linear, relu, selu, sigmoid, softplus, softsign, swish, tanh (c.f. FIGS. 2 to 4).
As to Claim 6, Bhattacharya teaches the limitations of Claim 3 and, wherein the first audio data, or one or more characteristic values based on the first audio data, is provided as a further input to the activation function code bank to further guide the selection of an activation function, Figure 9 and [0089] teaches conditional neural network is used to process input data at inference time according to the present techniques. As shown, Input-A (i.e., conditioning data) is optionally passed through the embedding network to obtain a low dimensional conditioning vector. For example, in case of personalized sound enhancement or speaker dependent automatic speech recognition (ASR) we would use speaker's enrolment data and would obtain a lower dimensional conditioning vector from that. Whereas, for the case of text-to-speech (TTS) conditioning vector would be gender (discrete), accent (continuous), speed of spoken output (continuous), etc, which are not required to be pre-processed for obtaining their low dimensional representations. Finally, the conditioning vector is then used to learn activation functions for specific conditions. These conditional activations are then used to condition the learned layers, which as depicted in the figure can be used together with any type of linear layers such as Dense, Cony or without non-linear layers such as LSTM/GRU.
As to Claim 10, Bhattacharya teaches the limitations of Claim 1 and wherein adapting the activation functions is at least partly based on the first audio data received via a microphone of the audio device, as 0010] teaches producing compact machine learning models suitable for deployment on resource-constrained apparatus, such as smartphones and Internet of Things ‘IoT’ devices, the devices implicitly receives audio data via microphone.
As to Claim 15, Bhattacharya teaches the limitations of Claim 1 and comprising a processor and a memory, wherein the audio device is configured for performing the method according to claim 1, ([0010] teaches producing compact machine learning models suitable for deployment on resource-constrained apparatus, such as smartphones and Internet of Things ‘IoT’ devices.
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.
1. Claims 2 and 7 are rejected under 35 U.S.C. 103 as being unpatentable over Bhattacharya et al. (US 2022/0375462A1), hereinafter “Bhattacharya’ in view of Kamkar-Parsi et. al. (US 2022/0201406A1), hereinafter “Kamkar-Parsi”.
As to Claim 2, Bhattacharya teaches the limitations of Claim 1 but does not explicitly teach wherein the method comprises: obtaining second audio data; determining one or more characteristic values based on the second audio data, wherein the one or more characteristic values are related to a sound environment; and based on the input and on the determined one or more characteristic values, adapting the activation function(s) of the one or more nodes of the neural network, while maintaining the topology of the neural network, thereby allowing the neural network to adapt in a computationally efficient way. However, training the neural network based on characteristics of a sound environment is well known in the art. However, Kamkar-Parsi in related field ( neural network) teaches as part of signal processing for hearing aids an artificial neural network is used and is implement in the hearing instrument. See at least abstract. Further, [0035] Kamkar-Parsi teaches the parameter for the signal processing: a voice activity, an activity of the user's own voice, a direction of a sound source, a speech detection, a recognition of a specific speaker, a classification of a listening situation, a characteristic variable for noise suppression, a characteristic variable for directional microphony. This means that the corresponding operation performs a voice activity detection (VAD), an OVD, a DOA, a detection (i.e. identification) of a specific speaker, a classification of the listening situation, a noise suppression, or directional microphony. It would have been obvious to one of ordinary skill in the art, to further provide a second input based on the parameters of the sound environment to serve the functionality of acoustic devices such as hearing aids where the ambient conditions are important to discriminate useful signals from interference.
As to Claim 7, Bhattacharya in view of Kamkar-Paris teaches the limitations of Claim 2, and, wherein the method comprises: determining one or more nodes of the neural network to be adapted based on the one or more characteristic values, Kamkar-Paris teaches on [0028] n ambient situation is detected, in particular on the basis of the input signal, and the topology is defined on the basis of the ambient situation. An ambient situation can be characterized on the one hand by an acoustic environment, so that a classification into standardized listening situations based on corresponding acoustic characteristics also defines the topology of the DNN. On the other hand, an ambient situation can also be characterized by a location (in particular inside/outside of a closed room) as well as by a movement of the user of the hearing instrument, which can be determined by one or more suitable sensors (acceleration sensor, GPS, etc.).
3. Claims 3 is rejected under 35 U.S.C. 103 as being unpatentable over Bhattacharya et al. (US 2022/0375462A1), hereinafter “ Bhattacharya’.
As to Claim 3, Bhattacharya teaches the limitations of Claim 1 and regarding the following : wherein the activation function is selected in a code bank comprising a plurality of activations functions, and the code bank optionally also comprising coefficients associated with activations functions, and/or bounding coefficients, Bhattacharya teaches on [0082] teaches the initial activation function may be non-linear functions to be applied to a linear layer of the network. Of course, it will be appreciated that the initial activation functions may include linear functions also. By way of example, at least one layer may be a Dense Linear layer, and the identified initial activation functions may be non-linear functions {elu, exponential, hard sigmoid, linear, relu, selu, sigmoid, softplus, softsign, swish, tanh (c.f. FIGS. 2 to 4). Bhattacharya does not explicitly teach a “ code bank “ for activation functions. However, allocating memory banks for storing coefficients for activation functions is well-known in the art, and it would have been obvious to one of ordinary skill in the art to allocate code banks to store coefficients such as elu, exponential, hard sigmoid, linear, relu, selu, sigmoid, softplus, softsign, swish, tanh (c.f. FIGS. 2 to 4) to retrieve them when needed.
Allowable Subject Matter
Claims 5, 8, 9, 11-14 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 SUNITA JOSHI whose telephone number is (571)270-7227. The examiner can normally be reached 8-3.
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, Duc Nguyen can be reached at 5712727503. 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.
/SUNITA JOSHI/Primary Examiner, Art Unit 2691
Prosecution Insights