中文摘要：

对神经电活动的研究,需要对记录的神经锋电位（spike）进行检测和分类。传统的锋电位检测与分类的过程受噪声影响很大,尤其是在锋电位分类的过程中,由于神经信号采集很难达到高的信噪比,因此研究较低信噪比下检测与分类的算法具有重要的意义。本文提供了一种运算简单的神经信号锋电位的检测与分类算法。针对锋电位的检测,本文使用非线性能量算子（NEO）来提高信噪比。在锋电位分类过程中,通过获取不同类别锋电位的模板构建匹配滤波器,对检出的锋电位在不同滤波器下进行滤波,确定锋电位对锋电位模板的相关性,然后使用非线性能量算子来加强相关性,从而确定分类。通过使用自主设计的大鼠神经信号双通道采集系统采集到仿真数据（数据由NSS-128的神经信号模拟器产生）,获得了可准确量度的标准数据,通过在这组数据上加不同程度的噪声,测试文章提到的算法。在检测方面,信噪比有约5倍的提升。分类方面相比单纯的匹配滤波,在一定的信噪比范围内,正确率由90%提升到100%。同时采集了部分实际神经数据,并用文中的检测方法进行了测试。

英文摘要：

Studying neural activity via electrical recording relies on the ability to detect and sort neural spikes.Traditional spike detection and sorting process is greatly affected by noise,especially in the case of spike sorting.Actual neural signal acquisition is very difficult to achieve high SNR,so detection and sorting algorithm for lower SNR signal is important.In this study we describe a simple algorithm for spike detection and sorting.In the spike detection algorithm,a nonlinear energy operator（NEO） is used to improve SNR.We construct matched filters based on different spike templates,and filter the detected spikes using different matched filters,then confirm the classification of the signal with the NEO operator through enhancing the correlation between the spike and the template.Using our own designed rat neural signal dual-channel data acquisition system（the data are generated from the NSS-128 neural signal simulator）,we obtained the standard data that can be measured precisely.The proposed algorithm was tested under different noise levels.Experimental results show that the SNR is increased by 5 times.Compared with matched filter method,the proposed algorithm improves the classification accuracy from 90% to 100%.Meanwhile,we also acquired some real neural data on which the algorithm was tested.