Room Impulse Response Reshaping-Based Underdetermined Blind Source Separation in a Reverberant Environment

Abstract

A room impulse response reshaping-based underdetermined blind source separation algorithm is designed to solve the signal separation problem in a reverberant environment. The main contribution is the use of room impulse response reshaping technology to eliminate the influence of audible echo for the reverberant environment, improving the quality of the received signals. A new mathematical model of time-frequency mixing signals is established to reduce the approximation error of model transformation caused by high reverberation. Furthermore, an improved expectation maximization algorithm is proposed for real-time update learning rules of model parameters, and then the sources are separated using the estimators provided by the improved expectation maximization algorithm. Experimental results demonstrate that the proposed algorithm achieves better separation performance to speech and music mixtures while maintaining much better robustness than the popular expectation maximization algorithms.

Keywords: Blind Source Separation, Underdetermined Mixtures, Room Impulse Response Reshaping, Reverberant Environment