Abstract: A method for clustering audio objects may involve identifying a plurality of audio objects, wherein each audio object of the plurality of audio objects is associated with respective metadata that indicates respective spatial position information and respective rendering metadata. The method may involve assigning audio objects of the plurality of audio objects to categories of rendering metadata of a plurality of categories of rendering metadata, wherein at least one category of rendering metadata comprises a plurality of types of rendering metadata to be preserved. The method may involve determining an allocation of a plurality of audio object clusters to each category of rendering metadata. The method may involve rendering audio objects of the plurality of audio objects to an allocated plurality of audio object clusters based on the metadata that indicates spatial position information and based on the assignments of the audio objects to the categories of rendering metadata.

Abstract: A method for performing denoising on audio signals is provided. In some implementations, the method involves determining an aggressiveness control parameter value that modulates a degree of speech preservation to be applied. In some implementations, the method involves obtaining a training set of training samples, a training sample having a noisy audio signal and a target denoising mask. In some implementations, the method involves training a machine learning model, wherein the trained machine learning model is usable to take, as an input, a noisy test audio signal and generate a corresponding denoised test audio signal, and wherein the aggressiveness control parameter value is used for: 1) generating a frequency domain representation of the noisy audio signals included in the training set; 2) modifying the target denoising masks; 3) determining an architecture of the machine learning model; or 4) determining a loss during training of the machine learning model.

Abstract: A computer-implemented method includes processing a time-series input signal using an encoder to produce an encoded representation, segmenting the encoded representation into a plurality of patches, applying a masking operation to a subset of the patches to produce a masked encoded representation, processing the masked encoded representation using a transformer to generate contextual features, processing the contextual features using a decoder to generate a predicted frequency-domain representation of the time-series input signal, and adjusting parameters of the encoder and parameters of the transformer to minimize a loss between the predicted frequency-domain representation and a reference frequency-domain representation derived from the time-series input signal.

Abstract: An audio processing system and method which calculates, based on spatial metadata of the audio object, a panning coefficient for each of the audio objects in relation to each of a plurality of predefined channel coverage zones. Converts the audio signal into submixes in relation to the predefined channel coverage zones based on the calculated panning coefficients and the audio objects. Each of the submixes indicating a sum of components of the plurality of the audio objects in relation to one of the predefined channel coverage zones. Generating a submix gain by applying an audio processing to each of the submix and controls an object gain applied to each of the audio objects. The object gain being as a function of the panning coefficients for each of the audio objects and the submix gains in relation to each of the predefined channel coverage zones.

Abstract: Systems, methods, and computer program products for audio processing based on convolutional neural network (CNN) are described. The CNN architecture may comprise a multi-scale input block and a multi-scale nested block. The multi-scale input block may be configured to receive input data and to generate a first downsampled input data set by downsampling the input data. The multi-scale nested block may comprise a first encoding layer configured to generate a first encoded data set by performing a convolution based on the input data. The multi-scale nested block may comprise a second encoding layer configured to generate a second encoded data set by performing a convolution based on the first downsampled input data set. Furthermore, the multi-scale nested block may comprise a first convolutional layer configured to generate a first output data set by upsampling the second encoded data set, concatenating the first encoded data set and the upsampled second encoded data set, and performing a convolution.

Abstract: Computer-implemented methods and devices for combined audio separation and classification are provided. An estimated separated signal is time gated based on a determination of an audio classifier of, at least in part, the original mix of signals before separation. Combined separation, classification, and time gating of both the estimated signal and a residual signal are also provided.

Abstract: A method comprises: obtaining softmask values for frequency bins of time-frequency tiles representing an audio signal; reducing, or expanding and limiting, the softmask values; and applying the reduced, or expanded and limited, softmask values to the frequency bins to create a time-frequency representation of an estimated target source. An alternative method comprises, for each time-frequency tile: obtaining softmask values; applying the softmask values to the frequency bins to create a time-frequency domain representation of an estimated target source; obtaining a panning parameter and a source concentration estimates for the target source; determining, using the panning parameter estimate and the softmask values, a magnitude for the time-frequency representation of the estimated target source; determining, using the panning parameter estimate and the source phase concentration estimate, a phase for the time-frequency representation of the estimated target source; and combining the magnitude and the phase.

Abstract: A method of audio content identification includes using a two-stage classifier. The first stage includes previously-existing classifiers and the second stage includes a new classifier. The outputs of the first and second stages calculated over different time periods are combined to generate a steering signal. The final classification results from a combination of the steering signal and the outputs of the first and second stages. In this manner, a new classifier may be added without disrupting existing classifiers.

Abstract: An audio processing system and method which calculates, based on spatial metadata of the audio object, a panning coefficient for each of the audio objects in relation to each of a plurality of predefined channel coverage zones. Converts the audio signal into submixes in relation to the predefined channel coverage zones based on the calculated panning coefficients and the audio objects. Each of the submixes indicating a sum of components of the plurality of the audio objects in relation to one of the predefined channel coverage zones. Generating a submix gain by applying an audio processing to each of the submix and controls an object gain applied to each of the audio objects. The object gain being as a function of the panning coefficients for each of the audio objects and the submix gains in relation to each of the predefined channel coverage zones.

Abstract: In an embodiment, a method comprises: transforming one or more frames of a two-channel time domain audio signal into a time-frequency domain representation including a plurality of time-frequency tiles, wherein the frequency domain of the time-frequency domain representation includes a plurality of frequency bins grouped into subbands. For each time-frequency tile, the method comprises: calculating spatial parameters and a level for the time-frequency tile; modifying the spatial parameters using shift and squeeze parameters; obtaining a softmask value for each frequency bin using the modified spatial parameters, the level and subband information; and applying the softmask values to the time-frequency tile to generate a modified time-frequency tile of an estimated audio source.

Abstract: The present disclosure relates to a method and system for processing audio for source separation. The method comprises obtaining an input audio signal (A) comprising at least two channels and processing the input audio signal (A) with a spatial cue based separation module (10) to obtain an intermediate audio signal (B). The spatial cue based separation module (10) is configured to determine a mixing parameter of the at least two channels of the input audio signal (A) and modify the channels, based on the mixing parameter, to obtain the intermediate audio signal (B). The method further comprises processing the intermediate audio signal (B) with a source cue based separation module (20) to generate an output audio signal (C), wherein the source cue based separation module (20) is configured to implement a neural network trained to predict a noise reduced output audio signal (C) given the intermediate audio signal (B).

Abstract: A method and system for separating a target audio source from a multi-channel audio input including N audio signals, N>=3. The N audio signals are combined into at least two unique signal pairs, and pairwise source separation is performed on each signal pair to generate at least two processed signal pairs, each processed signal pair including source separated versions of the audio signals in the signal pair. The at least two processed signal pairs are combined to form the target audio source having N target audio signals corresponding to the N audio signals.

Abstract: The present disclosure relates to a method and audio processing arrangement for extracting a target mid (and optionally a target side) audio signal from a stereo audio signal. The method comprises obtaining (S1) a plurality of consecutive time segments of the stereo audio signal and obtaining (S2), for each of a plurality of frequency bands of each time segment of the stereo audio signal, at least one of a target panning parameter (?) and a target phase difference parameter (?). The method further comprises extracting (S3), for each time segment and each frequency band, a partial mid signal representation (211, 212) based on at least one of the target panning parameter (?) and the target phase difference parameter (?) of each frequency band and forming (S4) the target mid audio signal (M) by combining the partial mid signal representations (211, 212) for each frequency band and time segment.

Abstract: Volume leveling of an audio signal using a volume leveling control signal. The method comprises determining a noise reliability ratio w(n) as a ratio of noise-like frames over all frames in a current time segment, determining a PGC noise confidence score XPGN(n) indicating a likelihood that professionally generated content, PGC, noise is present in the time segment, and determining, for the time segment, whether the noise reliability ratio is above a predetermined threshold. When the noise reliability ratio is above the predetermined threshold, the volume leveling control signal is updated based on the PGC noise confidence score, and when the noise reliability ratio is below the predetermined threshold, the volume leveling control signal is left unchanged. Volume leveling is improved by preventing boosting of e.g. phone-recorded environmental noise in UGC, while keeping original behavior for other types of content.

Abstract: Diffuse or spatially large audio objects may be identified for special processing. A decorrelation process may be performed on audio signals corresponding to the large audio objects to produce decorrelated large audio object audio signals. These decorrelated large audio object audio signals may be associated with object locations, which may be stationary or time-varying locations. For example, the decorrelated large audio object audio signals may be rendered to virtual or actual speaker locations. The output of such a rendering process may be input to a scene simplification process. The decorrelation, associating and/or scene simplification processes may be performed prior to a process of encoding the audio data.

Abstract: Volume leveler controller and controlling method are disclosed. In one embodiment, A volume leveler controller includes an audio content classifier for identifying the content type of an audio signal in real time; and an adjusting unit for adjusting a volume leveler in a continuous manner based on the content type as identified. The adjusting unit may configured to positively correlate the dynamic gain of the volume leveler with informative content types of the audio signal, and negatively correlate the dynamic gain of the volume leveler with interfering content types of the audio signal.

Abstract: The present disclosure relates to a method for designing a processor (20) and a computer implemented neural network. The method comprises obtaining input data and corresponding ground truth target data and providing the input data to a processor (20) for outputting a first prediction of target data given the input data. The method further comprises providing the latent variables output by a processor module (21: 1, 21: 2, . . . 21: n?1) to a supervisor module (22: 1, 22: 2, 22: 3, . . . 22: n?1) which outputs a second prediction of target data based on latent variables and determining a first and second loss measure by comparing the predictions of target data with the ground truth target data. The method further comprises training the processor (20) and the supervisor module (22: 1, 22: 2, 22: 3, . . . 22: n?1) based on the first and second loss measure and adjusting the processor by at least one of removing, replacing and adding a processor module.

Abstract: A method for performing denoising on audio signals is provided. In some implementations, the method involves determining an aggressiveness control parameter value that modulates a degree of speech preservation to be applied. In some implementations, the method involves obtaining a training set of training samples, a training sample having a noisy audio signal and a target denoising mask. In some implementations, the method involves training a machine learning model, wherein the trained machine learning model is usable to take, as an input, a noisy test audio signal and generate a corresponding denoised test audio signal, and wherein the aggressiveness control parameter value is used for: 1) generating a frequency domain representation of the noisy audio signals included in the training set: 2) modifying the target denoising masks: 3) determining an architecture of the machine learning model: or determining a loss during training of the machine learning model.

Abstract: Diffuse or spatially large audio objects may be identified for special processing. A decorrelation process may be performed on audio signals corresponding to the large audio objects to produce decorrelated large audio object audio signals. These decorrelated large audio object audio signals may be associated with object locations, which may be stationary or time-varying locations. For example, the decorrelated large audio object audio signals may be rendered to virtual or actual speaker locations. The output of such a rendering process may be input to a scene simplification process. The decorrelation, associating and/or scene simplification processes may be performed prior to a process of encoding the audio data.

Abstract: Systems and methods for preserving headphone rendering mode (HRM) in object clustering are described. In an embodiment, an object-based audio data processing system includes a processor configured to receive a plurality of audio objects, wherein an audio object of the plurality of audio objects is associated with respective object metadata that indicates respective spatial position information and an HRM; determine a plurality of cluster positions by applying an extended hybrid distance metric to a spatial coding algorithm to calculate a partial loudness for each of the audio objects; render the audio objects to the cluster positions to form a plurality of clusters by applying the extended hybrid distance metric to the spatial coding algorithm to calculate object-to-cluster gains; and transmit the clusters to a spatial reproduction system.