Multichannel Audio Coding with Eigen-Decomposition for Spatial Fidelity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing coding methods for ambisonic signals, such as multi-mono and parametric coding, result in spatial distortions and degradation of the sound scene due to the lack of consideration for channel correlations, leading to artifacts like phantom sound sources and diffuse noise.

Innovation Solution

A method involving the decomposition of covariance matrices into eigenvalues and eigenvectors, followed by quantization of these parameters in the Euler angle or quaternion domain, to optimize the coding rate and reduce spatialization distortions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If multi-mono or parametric coding methods are used for ambisonic signals, then the coding rate is reduced, but spatial distortions and degradation occur due to lack of channel correlation consideration

Engineering Contradiction:
Improvecoding rateVSAvoidspatial accuracy
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent transforms the covariance matrix parameters (eigenvalues and eigenvectors) into Euler angle parameters for coding. This parameter transformation allows the spatial correlation information to be efficiently represented with fewer bits while maintaining spatial accuracy, resolving the contradiction between coding rate reduction and spatial precision preservation

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent extracts the essential spatial correlation information from the covariance matrix by computing its eigenvalues and eigenvectors, and further extracts the rotational information as Euler angles. This extraction process separates the critical spatial parameters from the full covariance matrix, enabling efficient coding without losing spatial accuracy

Inventive Principle:
Principle #2Taking out (Extraction)

2Manufacturing precision

If covariance matrices are coded directly without decomposition, then spatial information is preserved, but the coding rate increases significantly

Engineering Contradiction:
Improvespatial image fidelityVSAvoidcoding rate
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential rotational information from the covariance matrix through eigen-decomposition and Euler angle transformation. This extraction eliminates redundant information while preserving the spatial correlation structure, achieving efficient coding with reduced bit rate

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter representation from the full covariance matrix elements to Euler angles derived from eigen-decomposition. This parameter transformation reduces the number of parameters to be coded while maintaining the spatial information, resolving the contradiction between fidelity and coding efficiency

Inventive Principle:
Principle #35Parameter changes

3Device complexity

If channel correlations are not considered in coding, then the coding process is simpler, but spatial distortions and artifacts occur

Engineering Contradiction:
Improvecoding process complexityVSAvoidspatial image quality
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent extracts the channel correlation information through covariance matrix computation and eigen-decomposition. By taking out this essential correlation information in the form of Euler angles, the method maintains spatial image quality while keeping the coding process manageable through systematic processing steps

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4172986B1Optimised coding of an item of information representative of a spatial image of a multichannel audio signal
Publication Date: 2025.08.13 ORANGE SA
  • EP4172986B1 patent drawingFigure 1~2
  • EP4172986B1 patent drawingFigure 3a~3b
  • EP4172986B1 patent drawingFigure 4

AI summary

The invention relates to a method for the optimised coding of a multichannel sound signal, comprising the following steps: coding at least one audio signal channel from the original multichannel signal; dividing the original multichannel signal into frequency sub-bands; determining one covariance matrix for each frequency sub-band, representative of a spatial image of the original multichannel signal; decomposing the predetermined covariance matrices into eigenvalues; coding by quantisation of the parameters from the decomposition into eigenvalues comprising both eigenvalues and eigenvectors. The invention also relates to a decoding method for decoding the parameters from the decomposition into eigenvalues of the covariance matrix of the original multichannel signal. It relates to coding and decoding devices implementing the respective methods.