基于FastICA的语音分离与图像分离  

Study on Speech Separation and Image Separation Based on FastICA

作　　者：田其冲[1] 郑卫国[1] 孙大雷[1] 

机构地区：[1]中国矿业大学计算机科学与技术学院,徐州221116

出　　处：《电脑编程技巧与维护》2009年第16期90-91,共2页Computer Programming Skills & Maintenance

摘　　要：独立分量分析(ICA)基于信号的高阶统计量,能从混合信号中分离出既具有统计独立性又具有非高斯性的源信号,在诸多ICA算法中,固定点算法(也称FastICA)以其收敛速度快、分离效果好被广泛应用于信号处理领域。在介绍ICA的基本模型与FastICA算法的原理后,分别对混合的语音信号与图像信号进行了分离实验,仿真结果表明FastICA应用于语音分离与图像分离,效果都很好。Independent Component Analysis （ICA） based on the higher-order statistics of signals, can separate source signals which are both statistically independent and non-Gaussian from the mixing signals. The fixed-point algorithm ,also called FastICA, has fast convergence rate and good separation result, so it can be widely used in the signal processing. In this paper, the basic model of ICA and the principle of the FastICA algorithm are introduced. And then the simulation experiments on separating the mixing speech signals and mixing image signals were made. Based on the experimental results, it comes to the conclusion that the FastICA algorithm has good separation performance when it applied to speech separation or image separation.

关 键 词：独立分量分析 固定点算法 语音分离 图像分离 

分 类 号：TN912.3[电子电信—通信与信息系统]