基于最小分散系数与旋转变换的EP信号自适应盲提取方法

Adaptive Blind Estimation of Evoked Potentials in EEG based on Minimum Dispersion Coefficient and Revolving Transform

诱发电位对于诊断神经系统损伤和病变具有重要的意义。传统的EP信号提取与分离方法中,通常认为EP信号中混入的EEG等噪声是高斯分布的。近年来一些研究表明了EEG信号具有一定的非高斯特性,而Alpha稳定分布可以更好地描述实际应用中所遇到的具有显著脉冲特性的EEG噪声。本研究提出了一种适用于EP信号分离提取的基于最小分散系数准则与旋转变换的算法,即通过分散系数的最小化,从而使估计误差的平均幅度达到最小,再利用Givens矩阵求解混矩阵。计算机模拟和分析表明,这种算法在分数低阶Alpha稳定分布背景噪声条件下,具有良好韧性,对EP信号可有效进行分离提取。

Evoked potentials have been widely used to diagnose the injury and pathological change in the central nervous system. Traditional EP analysis is developed under the condition that the background noises are with Gaussian distribution. Recent studies have revealed that there are non-Gaussian distribution noises in EEG. Alpha stable distribution is better for modeling the impulsive noises in EEG. In this paper, we proposed a new algorithm based on minimum dispersion coefficient and revolving transform. When the dispersion parameter was minimized, the mean amplitude of the estimated error approached its minimum. Consequently the Givens matrix was utilized to get solution of the mixed matrix. The simulation experiments showed that the proposed algorithm was more robust than the conventional algorithm.