On-line Algorithm for Extraction of Specific Signals with Temporal Structure

Abstract

Blind source separation techniques based on statistical independence criteria require a large number of data samples to estimate higher-order statistics. Thus, those techniques are not suitable to either on-line adaptive modeling. In this work we developed both an online and a batch algorithms for semi-blind extraction of a desired source signal with temporal structure from linear mixtures . Here, we do not assume that sources are statistically independent but we use an a priori information about the autocorrelation function of primary sources to extract the desired signal. Also, we develop an analytical framework to guarantee convergence of the online algorithm based on second-order statistics. Extensive computer simulations and real data applications confirm the validity and high performance of the proposed algorithms.