Blind Source Separation Based on Convolution Mixture Speech Signals

In this paper, in view of the convolution mixture speech signals which is more popular in real environment, we introduce a new technique for blind source separation of speech signals. Differing from most other major approaches to this problem, we focus on the temporal structure of the signals. The main idea is to apply the decorrelation method in the time-frequency domain proposed by Molgedey and Schuster in the time-frequency domain.. We show some results of experiments with both artificially controlled data and speech data recorded in the real environment.