Ambisonic Audio Coding via Quaternion Rotation Interpolation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current coders for ambisonic spatialized sound data face challenges in complexity, delay, and effective decorrelation between channels, particularly in the MPEG-H codec, which introduces sub-optimal linear interpolation and decomposition issues.

Innovation Solution

A coding process that involves calculating a covariance matrix for ambisonic channels, rotating clean vectors to ensure consistency, and interpolating matrices between frames to maintain decorrelation and smooth transitions, thereby reducing audible artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If linear interpolation is used in MPEG-H codec for ambisonic channels, then the decoding process is simplified, but audible artifacts and decorrelation issues arise

Engineering Contradiction:
Improvedecoding process complexityVSAvoidaudio quality
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent changes the parameter domain from linear interpolation to spherical linear interpolation (slerp) in the quaternion domain. By representing rotation matrices as quaternions and performing interpolation in this alternative parameter space, the method maintains mathematical consistency of rotation operations while achieving smooth transitions between frames, thereby resolving the contradiction between simplified decoding and audio quality.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent substitutes the linear interpolation mechanism with a quaternion-based spherical linear interpolation mechanism. This replacement changes the fundamental mathematical approach from vector space linear interpolation to rotation space quaternion interpolation, eliminating the decorrelation issues that arise from linear interpolation while maintaining computational feasibility.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If rotation matrices are not consistent between frames, then encoding flexibility is improved, but channel decorrelation deteriorates

Engineering Contradiction:
Improveencoding flexibilityVSAvoidchannel decorrelation
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent implements a feedback mechanism by computing the difference between consecutive rotation matrices and using this information to adaptively determine the number of interpolation subframes. The decoder uses the transmitted rotation parameters to reconstruct the same interpolation process, ensuring that both encoder and decoder maintain consistent channel decorrelation while preserving encoding flexibility through adaptive parameter transmission.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent performs preliminary computation of rotation matrix differences and determines the optimal number of interpolation subframes before the actual encoding process. By pre-calculating these parameters and transmitting them in the bitstream, the method ensures that decorrelation is maintained from the outset while allowing flexible adaptation to different audio scenarios.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If frame-based processing is used, then processing speed is improved, but temporal smoothness deteriorates due to audible clicks between frames

Engineering Contradiction:
Improveprocessing speedVSAvoidtemporal smoothness
Core Design Contradiction:
ProductivityVSStability of the object's composition

Solution Approach 1:

The patent segments each frame into multiple interpolation subframes when rotation matrix differences exceed a threshold. By dividing the frame processing into finer temporal segments with intermediate rotation steps, the method maintains the efficiency of frame-based processing while eliminating temporal discontinuities and audible clicks through gradual rotation transitions within each frame.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4498367A1Spatialized audio coding with interpolation and quantification of rotations
Publication Date: 2025.01.29 ORANGE SA
  • EP4498367A1 patent drawingFigure 1
  • EP4498367A1 patent drawingFigure 2
  • EP4498367A1 patent drawingFigure 3

AI summary

The invention relates to the compression coding of sound signals of N channels forming a succession in time of frames (t-1, t) of samples, the method comprising: - forming, from the channels for a current frame (t), a covariance matrix between channels and searching for eigenvectors of the covariance matrix to obtain an eigenvector matrix, - comparing the eigenvector matrix obtained for the current frame (t) to a rotation matrix obtained for a frame (t-1) preceding the current frame (t), and - permuting columns of the eigenvector matrix of the current frame (t) to ensure consistency with the rotation matrix of the previous frame (t-1).- test the eigenvector matrix to verify that it represents a rotation in an N-dimensional space and otherwise correct the eigenvector matrix until a rotation matrix is ​​obtained, for the current frame (t), and - apply said rotation matrix to the signals of the N channels before separate channel encoding of said signals, the process further comprising: - cutting the current frame into subframes, and - interpolating matrices to be applied to successive subframes from the matrix of the previous frame (t-1) to the matrix of the current frame (t).