A novel wavelet based ICA technique using Kurtosis

We consider the problem of blind audio source separation. A method to solve this problem is blind source separation (BSS) using independent component analysis (ICA). ICA exploits the non-Gaussianity of source in the mixtures. In this paper we propose a new wavelet based ICA method using Kurtosis for blind audio source separation. In this method, the observations are transformed into an adequate representation using wavelet packets decomposition and Kurtosis criterion. We consider instantaneous mixture of four sources. The results of performance measures show a considerable improvement when compared to FastICA and similar method.