摘要
基于人工神经网络和模板匹配,提出了对低信噪比神经元信号分类的方法.首先对待分类信号进行阈值检测,获得尖峰信号,对这些尖峰信号进行主成分分析,再选取主成分进行聚类,根据聚类结果,取对应的尖峰信号作为人工神经网络的训练样本.网络测试和结合模板匹配识别叠加信号的仿真结果表明了该方法的优越性.
A method to classify neural spikes under low signal-to-noise ratio was investigated by using artificial neural networks and template-matching. The spike events are extracted from the simulated multiunit spike train using peak detecting algorithm. Every spike event is analyzed by principal component analysis technique, then clustering analysis is applied to the principal components. Based on the result of clustering, the corresponding spike trains are used to train the network. The efficiency of the method is shown by the results of simulation.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2006年第5期852-855,共4页
Journal of Shanghai Jiaotong University
基金
国家自然科学基金(60375039)
上海市科委重点基金(02JC14008)资助项目