论文部分内容阅读
眼电伪迹和噪声是导致脑电信号低信噪比的重要原因。提出一种改进的基于少通道数的分块欠定盲源分离的滤波方法,通过分块的思想把非平稳的脑电信号变为平稳的分块信号,利用二阶欠定混合矩阵盲识别方法估计混合分离矩阵,然后通过基于最小均方误差的波速形成器提取源信号,最后通过得分准则自动去除噪声信号并重构信号。想象运动的真实脑电信号实验仿真结果表明分块欠定盲源分离方法能很好的恢复源信号并能有效地去除眼电等伪迹和噪声,提高了想象任务识别率。
EEG artifacts and noise are the important reasons for the low signal-to-noise ratio of EEG signals. An improved filtering method based on few-channel number-based unbundling blind source separation is proposed. The non-stationary EEG signal is changed into a stationary block signal by the idea of blocking. The second-order underdetermined mixed matrix is used to identify blindly The method estimates the mixed separation matrix, then extracts the source signal through the wave former based on the minimum mean square error, and finally automatically removes the noise signal and reconstructs the signal by using the scoring criterion. The experimental results of the real EEG signals of the imagination show that the block underdetermined blind source separation method can recover the source signals well and effectively remove artifacts such as electrooculogram and noise and improve the recognition rate of imaging tasks.