# US 2019/0318237 A1 · Systems, methods, and computer-readable media for improved real-time audio processing > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2019/0318237 A1 - Document ID: US20190318237A1 - Status: Published application - Kind: A1 - Application number: 15/952,329 - Filed: 2018-04-13 - Published: 2019-10-17 - Earliest claimed: 2018-04-13 - Assignees: Microsoft Technology Licensing, LLC - Applicants: Microsoft Technology Licensing, LLC - Inventors: Ivan Jelev TASHEV; Shuayb M ZARAR; Matthai PHILIPOSE; Jong HWAN KO - Web page: https://patentgrep.com/patent/US20190318237A1 ## CPC - G06N 3/08: Learning methods - G06F 3/165: {Management of the audio stream, e.g. setting of volume, audio stream path} - G06N 3/045: Combinations of networks - G06N 3/0495: Quantised networks; Sparse networks; Compressed networks - G06N 3/0499: Feedforward networks - G06N 3/09: Supervised learning - G10L 25/30: using neural networks - G10L 25/78: Detection of presence or absence of voice signals (switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10) - G06N 3/063: using electronic means - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract Systems, methods, and computer-readable storage devices are disclosed for improved real-time audio processing. One method including: constructing a deep neural network model, including a plurality of at least one-bit neurons, configured to output a predicted label of audio data, the plurality of at least one-bit neurons arranged in a plurality of layers, including at least one hidden layer, and being connected by a plurality of connections, each connection having at least a one-bit weight, wherein one or both of the plurality of at least one-bit neurons and the plurality of connections have a reduced bit precision; receiving a training data set, the training data set including audio data; training the deep neural network model using the training data set; and outputting a trained deep neural network model configured to output a predicted label of real-time audio data. ## Claims ### Claim 1 A computer-implemented method for improved real-time audio processing, the method comprising: constructing a deep neural network model, including a plurality of at least one-bit neurons, configured to output a predicted label of audio data, the plurality of at least one-bit neurons arranged in a plurality of layers, including at least one hidden layer, and being connected by a plurality of connections, each connection having at least a one-bit weight, wherein one or both of the plurality of at least one-bit neurons and the plurality of connections have a reduced bit precision; receiving a training data set, the training data set including audio data; training the deep neural network model using the training data set; and outputting a trained deep neural network model configured to output a predicted label of real-time audio data. ### Claim 2 (depends on 1) The method according to claim 1, wherein the deep neural network comprises one of one hidden layer, two hidden layers, and three hidden layers. ### Claim 3 (depends on 1) The method according to claim 1, wherein the connections include one-bit weights. ### Claim 4 (depends on 1) The method according to claim 1, wherein the plurality of at least one-bit neurons are one of one-bit neurons and two-bit neurons. ### Claim 5 (depends on 1) The method according to claim 1, wherein the at least one hidden layer includes less neurons than an input layer of the deep neural network and an output later of the deep neural network. ### Claim 6 (depends on 5) The method according to claim 5, wherein the at least one hidden layer may include one of 32, 128, and 512 neurons. ### Claim 7 (depends on 1) The method according to claim 1, wherein the received training data set includes audio data that has been previously scaled for precision. ### Claim 8 (depends on 1) The method according to claim 1, further comprising: assigning a first representative bit for each input value of audio data of the training data set based on a sign of each input value; calculating an average distance from a predetermined reference value; assigning a second representative bit for each input value of audio data of the training data set based on the calculated average distance; calculating a second average distance based on the calculated average distance; and assigning, for each input value, an approximate value as a reference of each section of the assigned first representative bit and second representative bit based on the second average distance. ### Claim 9 (depends on 1) The method according to claim 1, further comprising: receiving a test data set, the text data set including audio data with unseen noise; and evaluating the trained deep neural network using the received test data set. ### Claim 10 (depends on 1) The method according to claim 1, wherein the steps are iteratively or concurrently performed to output a plurality of trained deep neural network models, each configured to output a predicted label of real-time audio data, wherein each constructed deep neural network comprises at least one of a different number of neurons, a different number of neuron bits, a different number of layers, and a different number of weight bits. ### Claim 11 (depends on 1) The method according to claim 1, further comprising: receiving audio data, wherein audio data includes one or more audio frames represented in one of a time domain or a conversion domain; processing the audio data in segments using the trained deep neural network, wherein each segment includes one or more frames of the audio data. ### Claim 12 A system for improved real-time audio processing, the system including: a data storage device that stores instructions for improved real-time audio processing; and a processor configured to execute the instructions to perform a method including: constructing a deep neural network model, including a plurality of at least one-bit neurons, configured to output a predicted label of audio data, the plurality of at least one-bit neurons arranged in a plurality of layers, including at least one hidden layer, and being connected by a plurality of connections, each connection having at least a one-bit weight, wherein one or both of the plurality of at least one-bit neurons and the plurality of connections have a reduced bit precision; receiving a training data set, the training data set including audio data; training the deep neural network model using the training data set; and outputting a trained deep neural network model configured to output a predicted label of real-time audio data. ### Claim 13 (depends on 12) The system according to claim 12, wherein the deep neural network comprises one of one hidden layer, two hidden layers, and three hidden layers. ### Claim 14 (depends on 12) The system according to claim 12, wherein the connections include one-bit weights. ### Claim 15 (depends on 12) The system according to claim 12, wherein the plurality of at least one-bit neurons are one of one-bit neurons and two-bit neurons. ### Claim 16 (depends on 12) The system according to claim 12, wherein the at least one hidden layer includes less neurons than an input layer of the deep neural network and an output later of the deep neural network. ### Claim 17 (depends on 12) The system according to claim 12, wherein the received training data set includes audio data that has been previously scaled for precision. ### Claim 18 (depends on 12) The system according to claim 12, wherein the processor is further configured to execute the instructions to perform the method including: assigning a first representative bit for each input value of audio data of the training data set based on a sign of each input value; calculating an average distance from a predetermined reference value; assigning a second representative bit for each input value of audio data of the training data set based on the calculated average distance; calculating a second average distance based on the calculated average distance; and assigning, for each input value, an approximate value as a reference of each section of the assigned first representative bit and second representative bit based on the second average distance. ### Claim 19 A computer-readable storage device storing instructions that, when executed by a computer, cause the computer to perform a method for improved real-time audio processing, the method including: constructing a deep neural network model, including a plurality of at least one-bit neurons, configured to output a predicted label of audio data, the plurality of at least one-bit neurons arranged in a plurality of layers, including at least one hidden layer, and being connected by a plurality of connections, each connection having at least a one-bit weight, wherein one or both of the plurality of at least one-bit neurons and the plurality of connections have a reduced bit precision; receiving a training data set, the training data set including audio data; training the deep neural network model using the training data set; and outputting a trained deep neural network model configured to output a predicted label of real-time audio data. ### Claim 20 (depends on 19) The computer-readable storage device according to claim 19, wherein the method further comprises: assigning a first representative bit for each input value of audio data of the training data set based on a sign of each input value; calculating an average distance from a predetermined reference value; assigning a second representative bit for each input value of audio data of the training data set based on the calculated average distance; calculating a second average distance based on the calculated average distance; and assigning, for each input value, an approximate value as a reference of each section of the assigned first representative bit and second representative bit based on the second average distance.