# US 11,910,163 B2 · Signal processing device, system and method for processing audio signals > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,910,163 B2 - Document ID: US11910163B2 - Status: Granted - Kind: B2 - Application number: 17/425,248 - Filed: 2020-01-24 - Published: 2022-03-24 - Granted: 2024-02-20 - Earliest claimed: 2019-01-25 - Assignees: Sonova AG - Applicants: Sonova AG - Inventors: Peter Udo Diehl; Elias Sprengel - Web page: https://patentgrep.com/patent/US11910163B2 ## CPC - H04R 25/507: {implemented by neural network or fuzzy logic} - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/045: Combinations of networks - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/0475: Generative networks - G06N 3/08: Learning methods - G06N 3/088: Non-supervised learning, e.g. competitive learning - G06N 3/09: Supervised learning - G10L 21/0272: Voice signal separating - G10L 25/30: using neural networks - H04R 25/353: {Frequency, e.g. frequency shift or compression} - G06N 3/047: Probabilistic or stochastic networks - H04R 2225/43: Signal processing in hearing aids to enhance the speech intelligibility - H04R: LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; ELECTRIC HEARING AIDS; PUBLIC ADDRESS SYSTEMS (producing sounds with frequency not determined by supply frequency G10K) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract A signal processing device for processing audio signals is described. The signal processing device has an input interface for receiving an input signal and an output interface for outputting an output signal. Moreover, the signal processing device has at least one first neural network for conditioning the input signal and at least one second neural network for separating one or more audio signals from the input signal. The at least one first neural network and the at least one second neural network are arranged sequentially. ## Claims ### Claim 1 A signal processing device for processing audio signals, having: an input interface for receiving an input signal, at least one first neural network for conditioning the input signal, at least one second neural network for separating one or more audio signals from the conditioned input signal, and an output interface for outputting an output signal, wherein the output signal comprises the one or more audio signals separated from the conditioned input signal, wherein for the processing of the input signal the at least one first neural network and the at least one second neural network are arranged sequentially. ### Claim 2 (depends on 1) The signal processing device according to claim 1, characterized by a plurality of second neural networks, wherein each of the second neural networks is customized to a specific type of audio signals. ### Claim 3 (depends on 1) The signal processing device according to claim 1, characterized in that at least two second neural networks are used in parallel for the separation of the audio signals from the input signal. ### Claim 4 (depends on 1) The signal processing device according to claim 1, characterized in that the at least one second neural network used for separating the one or more audio signals is selectable from a plurality of second neural networks specializing in different types of audio signals. ### Claim 5 (depends on 1) The signal processing device according to claim 1, characterized in that the at least one first neural network and the at least one second neural network are part of a joint neural network. ### Claim 6 (depends on 1) The signal processing device according to claim 1, characterized by a user interface for receiving user inputs and/or for outputting information to a user. ### Claim 7 (depends on 1) The signal processing device according to claim 1, characterized by at least one data memory for storing known types of audio signals. ### Claim 8 (depends on 1) A system for processing audio signals, having: at least one signal processing device for processing audio signals according to claim 1, at least one recording device for recording the input signal, wherein the recording device is connected to the signal processing device via the input interface in a data-transmitting manner, and at least one playback device for playing back an output signal, wherein the playback device is connected to the signal processing device via the output interface in a data-transmitting manner. ### Claim 9 (depends on 8) The system according to claim 8, characterized by being a hearing device system. ### Claim 10 (depends on 8) The system according to claim 8, characterized in that the at least one signal processing device is in the form of a mobile phone, part of a mobile phone or a wireless microphone. ### Claim 11 (depends on 8) The system according to claim 8, characterized by a modular design. ### Claim 12 (depends on 1) A method for processing audio signals, having the steps of: providing a signal processing device according to claim 1, providing the input signal, supplying the input signal to the signal processing device via the input interface, conditioning the input signal by using the at least one first neural network, separating one or more audio signals from the conditioned input signal by using the at least one second neural network, determining a priority parameter for each of the audio signals, modulating each audio signal on the basis of the respective priority parameter, combining the audio signals to produce an output signal, outputting the output signal via the output interface. ### Claim 13 (depends on 12) The method according to claim 12, characterized in that the separated audio signals are classified. ### Claim 14 (depends on 12) The method according to claim 12, characterized in that a selection of the at least one second neural network from an available set of different second neural networks is made on the basis of user-specific data and/or already separated audio signals. ### Claim 15 (depends on 12) The method according to claim 12, characterized in that the determination of the priority parameters takes place asynchronously in relation to the further steps of the method. ### Claim 16 (depends on 12) The method according to claim 12, characterized in that the determination of the priority parameters takes place on the basis of user-specific data, preferences of the user and/or an information content of the respective audio signal. ### Claim 17 (depends on 12) The method according to claim 12, characterized in that the conditioning of the input signal and/or the separation of one or more audio signals from the conditioned input signal comprises one or more steps of classical audio processing. ### Claim 18 (depends on 12) The method according to claim 12, characterized in that the one or more audio signals separated from the input signal are enhanced. ### Claim 19 (depends on 12) The method according to claim 12, characterized in that the one or more audio signals separated from the input signal are denoised. ### Claim 20 (depends on 12) The method according to claim 12, characterized in that modulation of the one or more audio signals comprises amplification and/or frequency translation of at least parts of the one or more audio signal.