摘要
本文重点研究了瞬时混合模型下的语音信号分离算法,详细讨论了基于信息理论的独立分量分析算法,采用FastICA算法对多组语音信号进行了分离实验。仿真实验结果表明,该算法能正确的分离出原始语音信号,性能较好。同时也初步研究了基于统计学的JADE算法,将其与FastICA在加入噪声的条件下进行了性能比较仿真实验。实验结果表明,其稳定性优于JADE算法。
This paper focuses on the instantaneous mixture model of speech signal separation algorithm,a detailed discussion of information theory-based independent component analysis algorithm is given,FastICA algorithm is used to separate speech signal separation.Simulation results show that the algorithm is able to separate the original speech signal.Preliminary study of the JADE algorithm is also given.Joining with the noise,experimental results show that the JADE algorithm has a better performance than FastICA.
出处
《中山大学研究生学刊（自然科学与医学版）》
2011年第4期84-95,共12页
Journal of the Graduates Sun YAT-SEN University(Natural Sciences.Medicine)
