基于遗传算法的多通道癫痫脑电信号盲源分离

A Novel GA Based Blind Signal Separation of Multi-Channel EEG

目的:研究一种将心电噪声信号从脑电信号中分离出来的算法及其DSP硬件实现。方法:癫痫是一种中枢神经系统疾病,该病的诊断主要依靠脑电监测,但由于人体是一个复杂网络,临床采集到的脑电通常会混有其他噪声如心电干扰,这为后续的处理引入不可控制的误差。本文采用基于遗传算法的独立分量方法实现多通道脑电信号的盲源分离。结果:通过相关临床专家检验,认为该方法基本能够去除心电噪声,和参考心电信号对比具有一致性。结论:通过从北京某三甲医院癫痫中心采集到的患者脑电数据进行测试,对比试验表明,该方法是一种稳健高效的处理方法,符合并行运算的特点,整套算法可以移植到基于DSP的嵌入式系统架构上,具有一定的实用价值。

Objective： An analysis algorithm of GA based blind signal separation fi^om Multi-Channel EEG was designed in the article. Methods：The EEG （electroencephalogram） signal is a whole express way to show it＇s complicated electronic composition signal. It is a generally-accepted test method of epilepsy. An analysis algorithm of 24 lead EEG signal and it＇s embedded development system circuit method is discussed in the article. The method is based on genetic algorithms and fast ICA（FICA）. And here a novel GA process is designed to realize a high-speed and automatic estimation. Results：The experts of related fields considered that contrast with reference ECG signal ,the method is effective and accurate to remove ECG noise. Conclusions：The comparative experiments show the whole solution is a robust, effective and superior method to solve the EEG blind signal separation problem.