Dependent component analysis based on wavelet packet transform

作者：Zhang Yujie; Li Hongwei^*; Liu Zhihui

来源：Journal of Computational Information Systems, 2010, 6(13): 4275-4284.

• Component analysis
• Correlation
• Guasiantity
• Non-stationarities
• Wavelet packet transforms

This paper presents a blind source separation method which uses wavelet packet transform to obtain time-frequency information from a set of linear instantaneous mixtures of these sources. Unlike previously reported Time-Frequency blind source separation (TF BSS) methods, this method is based on wavelet packet transform and only requires the wavelet packet coefficients of each source to be non-negligible in one subband. It automatically determines these subbands using local second-order statistics parameters of the wavelet packet transform of the observed signals. Samples at these subbands can be used for the mixing matrix estimation. Different from classical independent component analysis methods, it is suited to non-stationary, dependence or Gaussian sources. Experimental results are provided to evaluate the performance of the proposed algorithm through comparing with other classical methods.

• 出版日期2010