语音信号识别基于盲源信号分离的实现  

Implementation of Speech Signal Recognition Base on Blind Source Signal Separation

作　　者：张敏[1] 于海燕[1] 武克斌[1] 

机构地区：[1]哈尔滨工业大学(威海),山东威海264209

出　　处：《现代电子技术》2011年第11期79-81,84,共4页Modern Electronics Technique

摘　　要：为了识别两路频谱混叠语音信号,多采用盲信号分离的方法。但是该方法在工程实践中实现较困难。因此给出了一种利用盲源信号分离的原理及特点的实现方法,具体说明了用FastICA算法在ADSP_BF533平台上实现盲源信号分离时的具体流程。该设计方案所需时间短,效率高,而且占用内存较少。The blind source separation（BSS） is usually used to distinguish two-channel sound signals of spectrum aliasing,but it is difficult to achieve in engineering practice.Therefore,a method which makes use of the principle and feature of blind source signal separation is introduced.The technological process to implement BSS with FastICA algorithm on ADSP_BF533 platform is described.The proposed scheme in this paper consumes less time,needs less memory and is high efficient.

关 键 词：盲信号分离 DSP FASTICA ADSP_BF533平台 

分 类 号：TN912.3[电子电信—通信与信息系统]