多维延时相关MUSIC方法:一种求解脑电逆问题的新方法

Multi-Dimensional Delay-Correlation MUSIC: A New Method to Extract Multi-Sources of EEGs

将经典的多信号分类算法 (MUSIC)用于研究脑电逆问题时存在两个问题 :对有色噪音敏感和不能识别相干源 .近年人们提出了利用延时相关、高阶累积量或假设已知噪音协方差来缓解有色噪音对算法的影响 .对于相干源 ,则有人提出了递归的多维MUSIC方法 .本文在这些工作的基础上建立了一种基于延时相关阵的、叠代的多维MU SIC算法 .仿真数据及实际脑电应用研究表明 ,该方法能在压制有色噪音的同时识别多个相干源 ,因而具有明显的意义 .

The reported studies on EEG inverse by multiple signal classification (MUSIC) show the classical MUSIC algorithm suffers from two shortcomings: be sensitive to a color noise and fail in identifying synchronously active sources. Recent studies reveal that the MUSIC-based algorithm may be improved in depressing spatial coherent noise if the classical zero delay correlation matrix is replaced by a non-zero delay-correlation matrix, or by a high-order cumulant matrix, or by incorporating known noise covariance matrix in the zero-delay correlation matrix. And the MUSIC algorithm can be extended to identify synchronous actives through a recursive strategy. In this work, an iterative, multi-dimensional and delay-correlation MUSIC is proposed, where the color noise is depressed by the non-zero delay correlation and the synchronous active sources are identified by the iterative multi-dimensional MUSIC search. Simulation and VEP data tests show a good reconstruction is obtained.