A Fast Adaptive Speech Extraction Method using Blind Source Separation for Audio Signal Processing

Abstract

The adaptive signal processing methods are used in several applications like channel estimation, Noise removal and extraction of signals also. The methods vary on time, frequency and statistical approach. In this paper, the source speech signals are separated using different methods like FastICA,PCA and kICA. Comparison of original signal and estimated signals are evaluated for different methods. The implementation was done in MATLAB. The spectrogram, Negentropy and Kurtosis waveforms are plotted for different methods.