FastICA与PCA结合的语音盲信号分离技术  

SPEECH BLIND SIGNAL SEPARATION TECHNOLOGY WITH FASTICA AND PCA COMBINATION

作　　者：赵志强[1] 颜学龙[1] 

机构地区：[1]桂林电子科技大学电子工程与自动化学院,广西桂林541004

出　　处：《计算机应用与软件》2012年第12期144-146,共3页Computer Applications and Software

基　　金：广西研究生教育创新计划项目(2010105950804M35)

摘　　要：由于语音盲信号分离中存在收敛速度慢的问题,在众多数学模型的基础上,采用FastICA和PCA结合的算法模型进行语音盲信号分离。在FastICA和PCA结合中,先通过PCA对混合信号处理,使得各道数据分析集中到高阶统计量上,然后采用快速ICA算法分离,从而使得收敛速度获得很大程度的提高。Since in speech blind signal separation there is the problem of slow convergence, based on many mathematical models, the algorithm model combining FastICA and PCA is employed to separate the speech blind signal. In such combination, first the mixed signal is processed with PCA and this makes the data analysis on all channels be centralised to higher-order statistics, then the FastICA algorithm is adopted for separation, thus the convergence speed is greatly improved.

关 键 词：快速ICA算法 主成分分析 源信号 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]