脑电诱发电位(EP)的单导少次提取一直是生物医学信号处理领域倍受关注的问题。本研究提出一种广义子空间法用于EP信号的单导少次提取问题中,实现对观测信号的滤波降噪得到EP信号在最小均方误差意义下的最佳估计。该算法的核心是首先利用投影矩阵将信号和噪声同时投影到系数空间,再根据观测信号和噪声的自相关矩阵得到系数加权矩阵,估计出信号的投影系数,最后利用重构矩阵进行重构得到期望的EP信号。仿真实验在不同初始信噪比条件下进行算法测试和性能分析,该算法较好地抑制了自发脑电的干扰,使信噪比获得了较大程度的提高。
The signal estimation of single channel brain evoked potential(EP) with few-trial is of great interest.In this paper,a generalized subspace approach was proposed to realize the optimum estimate of EP signals from the observable noisy signals with minimum mean square error(MSE) criterion.The underlying principle was to project the signals and noises into signal and noise coefficient subspace respectively by applying a projection matrix at first.Then,a coefficient weighting matrix was achieved based on the autocorrelation matrix of the noises and the noisy signals.Afterwards,the projection coefficients of the signal were estimated with the weighting matrix.EP signals were estimated with the reconstruction matrix.Simulation experiments were carried out to test and analyze the performance of the algorithm for the EP signals estimation in different signal-to-noise ratio(SNR) conditions.The interference of spontaneous electroencephalogram(EEG) was eliminated a lot with significantly improved SNR.The simulations results demonstrated the effectiveness and superior performance of the proposed method.