Audio Dereverberation via Eigenvalue Filter Coefficients

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current dereverberation and audio source separation techniques face limitations in complex acoustic scenarios, particularly in single-channel applications, and suffer from high computational complexity in multi-channel scenarios, failing to effectively separate audio signals in reverberant environments.

Innovation Solution

A signal processing apparatus and method that determines a filter coefficient matrix to make each output audio signal coherent with its own history and orthogonal to other audio source signals, using auto correlation and cross coherence matrices, applicable to both single-channel and multi-channel audio signals, and employing techniques like beamforming and eigenvalue decomposition for efficient dereverberation and source separation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If single-channel dereverberation techniques are used, then the system can be applied to single-channel audio signals, but the performance is limited in complex acoustic scenarios and cannot be generalized to multi-channel scenarios

Engineering Contradiction:
Improveapplicability to single-channel and multi-channel scenariosVSAvoidperformance in complex acoustic scenarios
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent develops a dereverberation system that universally handles both single-channel and multi-channel audio signals through a unified filter coefficient matrix approach. The system can adapt to different channel configurations while maintaining high performance in complex acoustic scenarios, resolving the contradiction between versatility and reliability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Reliability

If multi-channel dereverberation techniques are used, then audio source separation can be achieved in multi-channel scenarios, but the computational complexity becomes excessively high

Engineering Contradiction:
Improveaudio source separation capabilityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent transforms the audio signals into the frequency domain using Short-Time Fourier Transform (STFT), converting the time-domain dereververation problem into a frequency-domain filtering problem. This parameter transformation enables efficient computation of the filter coefficient matrix while maintaining effective audio source separation capability.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If conventional dereverberation techniques are used, then processing can be performed on audio signals, but speech intelligibility and sound quality deteriorate in reverberant environments

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidspeech intelligibility and sound quality
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent determines the filter coefficient matrix in advance based on the statistical properties of the audio signals and room impulse response characteristics. By pre-computing the optimal filter coefficients before actual dereverberation processing, the system achieves both high processing efficiency and improved speech intelligibility in reverberant environments.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3072129B1Signal processing apparatus, method and computer program for dereverberating a number of input audio signals
Publication Date: 2018.06.13 HUAWEI TECH CO LTD
  • EP3072129B1 patent drawingFigure 1
  • EP3072129B1 patent drawingFigure 2
  • EP3072129B1 patent drawingFigure 3

AI summary

The invention relates to a signal processing apparatus (100) for dereverberating a number of input audio signals, the signal processing apparatus (100) comprising a transformer (101) being configured to transform the number of input audio signals into a transformed domain to obtain input transformed coefficients, the input transformed coefficients being arranged to form an input transformed coefficient matrix, a filter coefficient determiner (103) being configured to determine filter coefficients upon the basis of eigenvalues resulting from the decomposition of an input auto-coherence matrix, the filter coefficients being arranged to form a filter coefficient matrix, a filter (105) being configured to convolve input transformed coefficients of the input transformed coefficient matrix by filter coefficients of the filter coefficient matrix to obtain output transformed coefficients, the output transformed coefficients being arranged to form an output transformed coefficient matrix, and an inverse transformer (107) being configured to inversely transform the output transformed coefficient matrix from the transformed domain to obtain a number of output audio signals.