Multi-Channel Sound Source Isolation Using Beamforming and ICA
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sound source separation technologies face challenges in efficiently isolating a speaker's voice from peripheral noise and reverb in multi-channel sound signals, leading to increased calculations and sound quality distortion.
Innovation Solution
An apparatus and method utilizing a microphone array with a signal processor performing Discrete Fourier Transform and Geometric Source Separation, followed by a post-processor that estimates noise variance and calculates a gain based on speech presence probability to separate speech signals, minimizing additional calculations and distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If additional signal processing based on voice estimation technologies is applied to improve separation performance, then speech clarity is improved, but the number of calculations increases
Solution Approach 1:
The patent applies beamforming to preliminarily estimate the positions of sound sources before applying ICA-based separation. By pre-identifying sound source locations and directions, the system reduces the search space and complexity for subsequent separation processing, achieving better speech clarity without proportionally increasing overall calculation burden
Solution Approach 2:
The patent divides the sound source separation process into distinct stages: beamforming for spatial filtering and sound source localization, followed by ICA for independent component separation. This segmentation allows each stage to focus on specific aspects of separation, improving overall effectiveness while managing computational complexity through modular processing
2Device complexity
If ICA alone is used for sound source separation, then the separation process is simple, but it cannot effectively separate sound sources in convolutive mixture environments
Solution Approach 1:
The patent merges beamforming technology with ICA-based BSS to create a hybrid separation system. Beamforming provides spatial filtering and sound source localization capabilities, while ICA handles the statistical independence-based separation. This combination overcomes the limitation of ICA alone in convolutive mixture environments while maintaining reasonable computational complexity
Solution Approach 2:
The patent introduces beamforming as an intermediary step between microphone signal acquisition and final sound source separation. The beamforming stage pre-processes the multi-channel signals to enhance spatial selectivity and provide initial sound source estimates, which then serve as inputs for the ICA-based separation stage, improving overall separation reliability
Data Source
AI summary
In an apparatus and method for isolating a multi-channel sound source, the probability of speaker presence calculated when noise of a sound source signal separated by GSS is estimated is used to calculate a gain. Thus, it is not necessary to additionally calculate the probability of speaker presence when calculating the gain, the speaker's voice signal can be easily and quickly separated from peripheral noise and reverb and distortion are minimized. As such, if several interference sound sources, each of which has directivity, and speakers are simultaneously present in a room with high reverb, a plurality of sound sources generated from several microphones can be separated from one another with low sound quality distortion, and the reverb can also be removed.