【摘要】 基于离散序列小波变换和主元分析,对低信噪比的神经元锋电位信号提出了新的分类方法.通过对原始信号进行尖峰检测,获得尖峰信号样本,对每个样本进行离散序列小波变换之后,再对小波变换系数进行主元分析,选取主元进行聚类,实现对原始信号的分类.将该方法应用于多电极细胞外记录的小鸡视网膜神经节细胞电活动信号分析,并据此推断出某电极附近的神经节细胞的个数.仿真结果表明,在低信噪比情况下,该方法比单纯通过小波变换进行分类的方法更有效.更多还原

【Abstract】 A new method was proposed to sort neural spikes under low signal-to-noise ratio by using discrete time wavelet transformation and principal component analysis. The events that represent spikes were extracted from the raw recording using a peak detecting algorithm, and a segment of fixed number of points was selected to straddle every event. Then the discrete time wavelet transformation was applied to analyze every spike event. After those coefficients of the discrete time wavelet transformation were analyzed by principal component analysis technique, a clustering analysis was applied to the principal components. Consequently the spikes were sorted according to the clustering result. The proposed method was also used to analyze the multi-channel extracellular recording of chicken’s isolated retina, from which the number of ganglion cells around the electrode was deduced. The efficiency of the method was well shown by the results of simulation. When the signal-to-noise ratio is low, the proposed method outperforms the method based on wavelet transformation.更多还原