基于递归神经网络结构的非平稳信号自适应盲分离  

Self-adaptive Blind Source Separation for Non-stationary Signals Based on Recursive Neural Network Structure

作　　者：苏伟生[1] 何文雪[2] 谢剑英[2] 

机构地区：[1]南京航空航天大学能源与动力学院,南京210016 [2]上海交通大学自动化所,上海200030

出　　处：《计算机工程》2004年第22期30-31,88,共3页Computer Engineering

基　　金：国家"863"计划基金资助项目(2001AA422420-02)

摘　　要：基于递归网络分离结构并利用时间相关的评价函数,针对二输入二输出盲信号分离问题,提出了一种非平稳信号的自适应盲分离算法。该算法计算量小,可根据输出信号能量大小有选择地更新分离系数,并可扩展到多输入多输出盲分离问题。仿真验证对声音等非平稳信号具有良好的分离效果。A double-input/double-output self-adaptive blind source separation (BSS) algorithm for non-stationary signals based on recursive structure and using time-related cost function is presented in this paper. This algorithm has low computing cost and could update selected separation parameters based on signal output energy which could be extended to multiple input/output blind source separation problem. Simulations have been made to validate the good separation performance of the algorithm.

关 键 词：信号处理 递归神经网络结构 自适应盲分离算法 仿真 

分 类 号：TN911.7[电子电信—通信与信息系统]