Power Spectrum Algorithm for Blind Source Separation
The phase differences of mixed signals are serious questions for blind source separation ( BSS ) and will weaken separation capacities of independent component analysis ( ICA ) algorithms. In order to solve this question, power spectrum BSS algorithm was proposed. Firstly, fundamental theories of BSS and fixed-point algorithm ( or FastICA algorithm ) were introduced. Owing to non-Gauss properties of autocorrelation processing and non-phase properties of power spectrum, an improved FastICA algorithm, called power spectrum ICA algorithm, was proposed. And its reasoning process was carried out. Then extended power spectrum algorithm was also studied. In simulations, validity of power spectrum BSS algorithm is argued by simulation mechanical signals. Theories and simulation results showed that power spectrum BSS algorithm is an effective method to solve phase differences of mixed signals for blind source separation.
blind source separation independent component analysis FastICA algorithm power spectrum
Wenfeng Wu Xiaohu Chen Xunjia Su Chunjiang Yao Xuping Wang
Xian Research Inst. of Hi-Tech Hongqing Town, Xi an, P. RXhina, 710025
国际会议
长春
英文
219-222
2010-08-24（万方平台首次上网日期，不代表论文的发表时间）