A Blind Source Separation Algorithm Based on Whitening and Non-linear Decorrelation
Zhijun Li, Jianping An, Лэй Сун, Miao Yang
Abstract
Zhijun Li, Jianping An, Лэй Сун, Miao Yang
Abstract
Blind source separation (BSS) methods become more attractive targets in neural network and signal processing literature recently. And independent component analysis (ICA) methods are one kind of widely used and important solutions to blind source separation problems. Whitening is a very useful preparation step for blind separation of source and meanwhile non-linear decorrelation acts as a basic estimation theory upon ICA methods. Then an on-line real-time blind source separation algorithm based on the whitening and non-linear decorrelation is proposed in this paper. Performance of the proposed method is simulated and analyzed and then the results show that our proposed algorithm has better convergence speed and steady-state error than the normalized EASI, nature gradient and whitening-only methods.
OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Blind source separation (BSS) methods become more attractive targets in neural network and signal processing literature recently. And independent component analysis (ICA) methods are one kind of widely used and important solutions to blind source separation problems. Whitening is a very useful preparation step for blind separation of source and meanwhile non-linear decorrelation acts as a basic estimation theory upon ICA methods. Then an on-line real-time blind source separation algorithm based on the whitening and non-linear decorrelation is proposed in this paper. Performance of the proposed method is simulated and analyzed and then the results show that our proposed algorithm has better convergence speed and steady-state error than the normalized EASI, nature gradient and whitening-only methods.
Key concepts: Decorrelation, Blind signal separation, Independent component analysis, Computer science, Algorithm, Convergence (economics), Source separation, Separation (statistics)