脑电逆问题多信号分类算法的计算机仿真研究

根据头表观测电位反演脑电源的空域位置 ,和它在时域中的演化过程是脑电研究中的一个重要方面。本文针对三层同心球头模型 ,首次采用归一化模糊指数反映反演结果的空域模糊程度 ,用奇异值比率反映多信号分类算法 (MUSIC)的信噪空间分离程度 ,并利用先进的总体最小二乘法反演脑电源的时域演化过程 ,实现了从反演结果的空域分布模糊程度、时域过程的重建精度和信噪空间分离程度三个方面对MUSIC算法的计算机仿真研究。研究内容包括不同信噪比的观测记录和不同相关性的源组合 ,结果显示了该算法对高斯白噪的稳健性和对相干源的敏感性 ,为进一步把该方法用于实践提供了依据。

A COMPUTER SIMULATION STUDY OF MUSIC ALGORITHM FOR EEG INVERSE PROBLEM

By using computer simulation, effectiveness of multiple signal classification (MUSIC) was evaluated. The head model used was the concentric 3 spheres conductor model. The EEG time course was reconstructed by the Total Least Squares (TLS). Two new parameters were proposed: Normalized Blurring Index (NBI) was used to indicate the spatial blurring level and Singular Value Ratio (SVR) was used to show the effectiveness of the signal noise subspace decomposition in the MUSIC algorithm. The general Relative Error (RE) and Correlation Coefficient (CC) were used to express the temporal reconstruction precision. It was simulated by using the different level of Gaussian white noise and the different correlation sources that the spatial blurring seriousness, the temporal reconstruction precision and the ability to decompose the signal and the noise subspace of the MUSIC algorithm. The results showed that the MUSIC algorithm was robust to the Gaussian white noise and sensitive to the sources correlation.