神经元锋电位检测中大幅值干扰的去除.docVIP

 神经元锋电位检测中大幅值干扰的去除# 万红，刘新玉* （郑州大学电气工程学院，郑州 450001） 5 10 15 20 25 30 35 40 摘要：神经元锋电位（spike）的有效提取在神经信号处理中具有十分重要的作用。然而由 于神经元锋电位检测信号是一种非平稳宽带高频小幅值信号，易受噪声影响而产生畸变，在 spike 检测中经常会将一些大幅值干扰误检为 spike，降低了 spike 后续分析的准确性和可靠 性。为了有效去除检测后 spike 中混杂的噪声干扰，提出了一种基于相关距离的 k-近邻方法 CDkNN（correlation distance based k-nearest neighbor），用于识别 spike 检测中误检的大幅值 噪声信号，并利用仿真实验和实测数据对算法的有效性进行了验证。仿真结果显示当 spike 的信噪比大于 0dB 时，识别正确率达到了 90%以上，对于提高 spike 后续分析结果的稳定性 和可靠性具有较高的实用价值。 关键词：生物信号处理；锋电位；相关距离；k-近邻；噪声 中图分类号：R318.4 Removing Interference Events in Neuronal Spikes Detection WAN Hong1, LIU Xinyu2 (1. Electrical Engineering School, Zhengzhou University, Zhengzhou 450001; 2. Electrical Engineering School, Zhengzhou University 450001) Abstract: Spike detection of single neurons form extracellular recordings are crucial for investigating the mechanisms of neural information processing in neuron system. However, because spikes data are a non-stationary, high- frequency and small amplitude signals, they are highly susceptible to noise, and also produce distortion, so it is often false detect some large amplitude noises as spikes in spike detection to reduce the accuracy and reliability of spike processing after detection. To improce the spike detection accuracy, a new algorithm based on CDkNN (correlation distance based k-nearest neighbor) was proposed to remove exactly noise events in spikes after detection. The effectiveness of CDkNN was verified by simulation and real data in V1. The simulation results show that the recognition accuracy of CDkNN above 90% when signal-to-noise ratio more than 0dB. Its high practical value for enhancing the accuracy and reliability of spikes follow-up result. Key words: biosignal processing; spike; correlation distance; k-nearest neighbor; noise 0 引言 锋电位（spike）活动的皮层内微电极胞外记录是深入探究大脑神经信息处理机制的基 本方式之一[1]。然而由于电极附近多个神经元活动的同时存在，往往难以实现单个电极对单 个神经元活动的记录。因此，需要通过对电极记录数据的分析处理进行神经元放电信号的检 测，也即需要将神经元 spike 活动从背景噪声中提取出来，并在这个基础上甄别出不同神经 元活动。在神经电生理实验中，微电极除了记录神经元放电活动之外，还会记录到肌电活动 或电磁干扰等噪声信号，尤其是