Abstract: Based on computational auditory scene analysis (CASA), this paper proposes an improved algorithm for monaural speech enhancement. In the proposed algorithm, both effective energy of target speech and crosschannel correlation are chosen as extracted feature. Moreover, this algorithm improves the threshold selection on energy spectrum and crosschannel correlation feature. Under the condition of low SNR with 5 different noises, the experimental results show that the proposed algorithm can raise the output SNR by 9.32 dB averagely, and attenuates noise effectively.