Frame Based Single Channel Speech Separation using Sumary Autocorrelation Function

Abstract

Single channel speech separation system is widely used in many applications. Pre-processing stage of Automatic speech recognition system, telecommunication system and the hearing aid design require the speech separation system to enhance the speech. This paper proposes a separation system that separates the dominant speech from the noisy environment, based on summary autocorrelation function (SACF) analysis pitch range estimation in the modulation frequency domain. Performance evaluation of the proposed system shows a better response compared to the existing methods.