A Single-Channel Blind Separation Method based on adaptive Morlet wavelet
Yan‐Feng Li, Yanhua Jin, Dake Liao, Hanjun Kou · 2021
The methods which are based on wavelet transform are often adopted to solve the single-channel blind separation, and the parameters of wavelets are related to the performance of these algorithms. However, these parameters rely on artificial adjustment, this greatly increases the difficulty of signal separation. In order to optimize the parameters to improve the performance of separation, a new method based on adaptive Morlet wavelet is proposed. First, the criterion of minimum Shannon entropy is utilized to optimize the parameters of wavelets. By using multi-scale wavelet decomposition based on optimized wavelet on single-channel signal, the original mixed signal can be expanded to a multi-dimensional signal without any prior knowledge, finally the noncircular fast independent component analysis (nc-FastICA) algorithm is used to realize its blind separation. The simulation result confirmed that the proposed method valid and feasible.