METHOD TO COMBINE CASA AND SOFT MASK FOR SINGLE-CHANNEL SPEECH SEPARATION

摘要

PURPOSE: A single-channel speech separation method using CASA and a soft mask algorithm are provided to correct voice loss in a clear input signal by applying a sound source separation mask based on the estimation of an SNR for each channel. CONSTITUTION: Before the separation of a sound source, an SNR(Signal to Noise Ratio) of an input voice signal is estimated(S110). Through the use of CASA(Computational Auditory Scene Analysis), the voice signal is firstly separated(S120). By using a soft mask, the voice signal is secondly separated(S130). After separating the sound source, the SNR of the voice signal is acquired(S140). The comparison and combination of SNR is performed(S150).