基于小波变换和盲源分离算法提取事件相关电位

阐述了用小波分解和盲源分离(blind source separation,BSS)算法结合来去除噪声和干扰提取事件相关电位(event-related potential,ERP).采用小波变换分解ERP,抽取出不同频带的细节信息;由小波系数判断选择多个尺度的子带信号,将它们分别与原始ERP组合进行盲分离,方法是极大化信号时间上的可预测性;将分离的结果进一步叠加平均.两类ERP仿真实验结果表明,本文算法提取出的ERP主要成分波明显,易于辨识,信噪比比较单独运用盲分离算法提取出的结果要好.在应用实例中,有效地增强了ERP的μ波.该算法优点在于减少了刺激次数和波形失真,参数变化范围小,在临床上有很好的应用前景.

Extraction of Event-Related Potential Based on Wavelet Transformation and Blind Source Separation

Blind source separation (BSS) combined with wavelet transformation is proposed to remove noises and artifacts, and extract event-related potential (ERP). First, ERP is decomposed by wavelet transformation, the detail signals of different frequency are extracted; Then multi-scale signals are selected by wavelet coefficients, they are differently combined with origin ERP to perform BSS,and the principle is maximizing a measure of temporal predictability for each recovered signal; Finally, the separated results are averaged ulteriorly.Simulation experiment results show this method can extract the distinct characteristic components of ERP.The ERP extracted by this method is distinguished easilier, SNR is higher compared with the results performed by BSS. In application example, mu rhythm of ERP is reinforced effectively. The method decreases stimulus time, distortion, and the parameter range,so it will has good future in clinic application.