【摘要】 目的大脑神经元胞外单细胞动作电位(即锋电位)的检测与分类,是研究神经系统处理信息机制的关键。常用方法是实验完成后对记录到的数据进行离线检测与分类,然而当需要在短时完成大量数据的处理或无线传输时,则需实现锋电位的在线检测与分类。方法为实现在线分类,本文在利用主成分分析法(principal component analysis,PCA)和K均值分类法对一定量数据进行预分类的基础上,提出使用PCA结合Fisher判别分析的方法,并与基于距离的模板匹配法、BP神经网络分类法进行了分类效果和算法复杂度的比较。结果仿真结果表明,该方法相对于其它两种方法在分类效果和算法复杂度上都具有一定的优势。结论此方法是实现锋电位在线分类的不错选择。更多还原

【Abstract】 Objective The detection and separation of neuron action potentials(spikes) is a key of information processing mechanism in the neural system research.Offline spike sorting approaches are currently used to detect and sort action potentials after experiments.Online sorting and detecting algorithms are required when researchers need to analyze or wireless transmit a large number of recordings in a short time.Methods To realize online spike sorting,we propose an algorithm combining Fisher discriminant analysis with principal component analysis(PCA) after performing a preprocessing with PCA and K-means.We test this algorithm and compare the efficiency and complexity with the other two algorithms;template matching based on distance and back propagation neural networks.Results The simulation results indicate that this method has an advantage over the other two methods in both efficiency and complexity.Conclusions This online spike sorting algorithm is effective.更多还原