Abstract: Computer-implemented methods for training a neural network, as well as for implementing audio encoders and decoders via trained neural networks, are provided. The neural network may receive an input audio signal, generate an encoded audio signal and decode the encoded audio signal. A loss function generating module may receive the decoded audio signal and a ground truth audio signal, and may generate a loss function value corresponding to the decoded audio signal. Generating the loss function value may involve applying a psychoacoustic model. The neural network may be trained based on the loss function value. The training may involve updating at least one weight of the neural network.

Abstract: Computer-implemented methods for training a neural network, as well as for implementing audio encoders and decoders via trained neural networks, are provided. The neural network may receive an input audio signal, generate an encoded audio signal and decode the encoded audio signal. A loss function generating module may receive the decoded audio signal and a ground truth audio signal, and may generate a loss function value corresponding to the decoded audio signal. Generating the loss function value may involve applying a psychoacoustic model. The neural network may be trained based on the loss function value. The training may involve updating at least one weight of the neural network.

Abstract: Described herein is a method of processing an audio signal using a neural network or using a first and a second neural network. Described is further a method of training said neural network or of jointly training a set of said first and said second neural network. Moreover, described is a method of obtaining and transmitting a latent feature space representation of a perceptual domain audio signal using a neural network and a method of obtaining an audio signal from a latent feature space representation of a perceptual domain audio signal using a neural network. Described are also respective apparatuses and computer program products.

Abstract: A system for suppressing noise and enhancing speech and a related method are disclosed. The system trains a neural network model that takes banded energies corresponding to an original noisy waveform and produces a speech value indicating the amount of speech present in each band at each frame. The neural model comprises a feature extraction block that implements some lookahead. The feature extraction block is followed by an encoder with steady down-sampling along the frequency domain forming a contracting path. The encoder is followed by a corresponding decoder with steady up-sampling along the frequency domain forming an expanding path. The decoder receives scaled output feature maps from the encoder at a corresponding level. The decoder is followed by a classification block that generates a speech value indicating an amount of speech present for each frequency band of the plurality of frequency bands at each frame of the plurality of frames.

Abstract: Computer-implemented methods for training a neural network, as well as for implementing audio encoders and decoders via trained neural networks, are provided. The neural network may receive an input audio signal, generate an encoded audio signal and decode the encoded audio signal. A loss function generating module may receive the decoded audio signal and a ground truth audio signal, and may generate a loss function value corresponding to the decoded audio signal. Generating the loss function value may involve applying a psychoacoustic model. The neural network may be trained based on the loss function value. The training may involve updating at least one weight of the neural network.

Abstract: Described herein is a method of waveform decoding, the method including the steps of: (a) receiving, by a waveform decoder, a bitstream including a finite bitrate representation of a source signal; (b) waveform decoding the finite bitrate representation of the source signal to obtain a waveform approximation of the source signal; (c) providing the waveform approximation of the source signal to a generative model that implements a probability density function, to obtain a probability distribution for a reconstructed signal of the source signal; and (d) generating the reconstructed signal of the source signal based on the probability distribution. Described are further a method and system for waveform coding and a method of training a generative model.

Abstract: Methods for detecting whether a rendered version of a specified seamless connection (“SSC”) at a connection point between two audio segment sequences results in an audible discontinuity, and methods for analyzing at least one SSC between audio segment sequences to determine whether a renderable version of each SSC would have an audible discontinuity at the connection point when rendered, and in appropriate cases, for a SSC having a renderable version which is determined to have an audible discontinuity when rendered, correcting at least one audio segment of at least one segment sequence to be connected in accordance with the SSC in an effort to ensure that rendering of the SSC will result in seamless connection without an audible discontinuity. Other aspects are editing systems configured to implement any of the methods, and storage media and rendering systems which store audio data generated in accordance with any of the methods.

Abstract: Computer-implemented methods for training a neural network, as well as for implementing audio encoders and decoders via trained neural networks, are provided. The neural network may receive an input audio signal, generate an encoded audio signal and decode the encoded audio signal. A loss function generating module may receive the decoded audio signal and a ground truth audio signal, and may generate a loss function value corresponding to the decoded audio signal. Generating the loss function value may involve applying a psychoacoustic model. The neural network may be trained based on the loss function value. The training may involve updating at least one weight of the neural network.

Abstract: A method for performing inter-channel encoding of a multi-channel audio signal comprising channel signals for N channels, with N being an integer, with N>1, is described. The method comprises determining a basic graph comprising the N channels as nodes and comprising directed edges between at least some of the N channels. Furthermore, the method comprises determining an inter-channel coding graph from the basic graph, such that the inter-channel coding graph is a directed acyclic graph, and such that a cumulated a cumulated cost of the signals of the nodes of the inter-channel coding graph is reduced.

Abstract: A method for performing inter-channel encoding of a multi-channel audio signal comprising channel signals for N channels, with N being an integer, with N>1, is described. The method comprises determining a basic graph comprising the N channels as nodes and comprising directed edges between at least some of the N channels. Furthermore, the method comprises determining an inter-channel coding graph from the basic graph, such that the inter-channel coding graph is a directed acyclic graph, and such that a cumulated a cumulated cost of the signals of the nodes of the inter-channel coding graph is reduced.

Abstract: A method of encoding adaptive audio, comprising receiving N objects and associated spatial metadata that describes the continuing motion of these objects, and partitioning the audio into segments based on the spatial metadata. The method encodes adaptive audio having objects and channel beds by capturing a continuing motion of a number N objects in a time-varying matrix trajectory comprising a sequence of matrices, coding coefficients of the time-varying matrix trajectory in spatial metadata to be transmitted via a high-definition audio format for rendering the adaptive audio through a number M output channels, and segmenting the sequence of matrices into a plurality of sub-segments based on the spatial metadata, wherein the plurality of sub-segments are configured to facilitate coding of one or more characteristics of the adaptive audio.

Abstract: Methods for detecting whether a rendered version of a specified seamless connection (“SSC”) at a connection point between two audio segment sequences results in an audible discontinuity, and methods for analyzing at least one SSC between audio segment sequences to determine whether a renderable version of each SSC would have an audible discontinuity at the connection point when rendered, and in appropriate cases, for a SSC having a renderable version which is determined to have an audible discontinuity when rendered, correcting at least one audio segment of at least one segment sequence to be connected in accordance with the SSC in an effort to ensure that rendering of the SSC will result in seamless connection without an audible discontinuity. Other aspects are editing systems configured to implement any of the methods, and storage media and rendering systems which store audio data generated in accordance with any of the methods.

Abstract: A method of encoding adaptive audio, comprising receiving N objects and associated spatial metadata that describes the continuing motion of these objects, and partitioning the audio into segments based on the spatial metadata. The method encodes adaptive audio having objects and channel beds by capturing a continuing motion of a number N objects in a time-varying matrix trajectory comprising a sequence of matrices, coding coefficients of the time-varying matrix trajectory in spatial metadata to be transmitted via a high-definition audio format for rendering the adaptive audio through a number M output channels, and segmenting the sequence of matrices into a plurality of sub-segments based on the spatial metadata, wherein the plurality of sub segments are configured to facilitate coding of one or more characteristics of the adaptive audio.