基于小波和卡尔曼平滑的事件相关电位单次提取

本研究提出一种事件相关电位单次提取方法,可有效减少实验次数,并可探索实验之间ERP的变异性。此方法基于小波和卡尔曼平滑,首先利用小波变换考察ERP平均信号的时频特性,根据ERP不同分量出现的时间位置,在不同尺度上选取特定的单次实验ERP小波系数构成观测向量,其为真实ERP小波系数状态向量与噪声之和,然后对观测向量进行卡尔曼平滑,最后对卡尔曼平滑后的小波系数进行小波重构,得到单次提取的ERP信号。仿真实验表明,基于小波和卡尔曼平滑的方法不仅信噪比提高约16～18 dB,优于30次叠加平均、简单小波方法和基于高斯基函数的卡尔曼滤波方法,还可以跟踪ERP的幅度趋势变异性。与基于高斯基函数的卡尔曼滤波方法相比,所提方法降低了计算量。真实脑电ERP提取实验表明本方法较好地从单次记录中提取出了事件相关电位,并可解释ERP因适应和应激引起的趋势变异性。

Single Trial Event Related Potential Extraction Based on Wavelet and Kalman Smoother Method

A single trial extraction method of event-related potential(ERP) was proposed.In this wavelet and Kalman smoother based method,wavelet transform was first applied to the averaged ERP to investigate its time-frequency characteristics.According to the time position of different components of ERP,the specific wavelet coefficients of single trial ERP on different scales were chosen to construct the measurement vector,which was filtered by Kalman smoother in the next step.Eventually,the disposed coefficients were reconstructed and the single trial ERP signal was obtained.The simulation experiments showed that the wavelet and Kalman smoother based method improved SNR about 16-18dB,and was better than 30 trial ensemble averaging method and the simple wavelet method.The proposed method could also track the ERP amplitude trend-like variability.The real dataset experiment showed that the method preferably extracted the ERP signal concealed in the ongoing background EEG and habituation or sensitization could explain the trend-like variability.