基于PCA和改进K均值算法的动作电位分类  

Action Potential Classification Based on PCA and Improved K-means Algorithm

作　　者：师黎[1] 杨振兴[1] 王治忠[1] 王岩[1] 

机构地区：[1]郑州大学电气工程学院,郑州450001

出　　处：《计算机工程》2011年第16期182-184,187,共4页Computer Engineering

基　　金：国家自然科学基金资助项目(60841004;60971110)

摘　　要：微电极阵列记录的神经元信号往往是电极临近区域数个神经元的动作电位信号以及大量背景噪声的混叠,研究神经系统的信息处理机制以及神经编码、解码机理需了解相关每个神经元的动作电位,因此需从记录信号中分离出每个神经元的动作电位。基于此,提出基于主元分析(PCA)和改进K均值相结合的动作电位分类方法。该方法采用PCA提取动作电位特征,使用改进K均值算法实现动作电位分类。实验结果表明,该方法降低了动作电位的特征维数以及K均值算法对初始分类重心的依赖,提高动作电位分类结果的正确率及稳定性。尤其是在处理低信噪比信号时,分类正确率仍能达到理想水平。Neural signal recorded by the microelectrode array is often the mixture which is composed of action potentials of several neurons near the electrodes and the background noises.Researches on the nervous system information processing mechanism and neural coding and decoding mechanism need know every related neuron＇s action potential.Therefore,every neuron＇s action potential is essential to be separated from the recorded signal.This paper proposes a method based on Principal Component Analysis（PCA） combined with improved K-means for action potential classification.The action potentials＇ features are extracted by PCA,the action potential classification is implemented by the improved K-means algorithm.Experimental results show that the method brings down action potential＇s feature dimensions and dependence of the initial classification center for the K-means algorithm,and increases the accuracy and stability of the classification results.Particularly,when processing the low Signal to Noise Ratio（SNR） signals,it can also achieve an expected purpose.

关 键 词：微电极阵列 主元分析 特征提取 改进K均值 动作电位分类 

分 类 号：R318.04[医药卫生—生物医学工程] TP391.4[医药卫生—基础医学]