论文部分内容阅读
由于大脑信息编码的稀疏特性,微电极阵列记录的神经元集群信号中包含大量的噪声和冗余信息,这降低了运动意图解码的稳定性和精确度。针对这一问题,本文将偏最小二乘(PLS)特征提取应用于神经元集群解码中,首先采用PLS提取神经元集群锋电位发放特征,然后用支持向量机(SVM)对提取的特征进行分类,解码得到运动意图。采集三组鸽子十字迷宫转向实验中的大脑神经元集群信号进行解码,结果表明,PLS结合分类模型的解码方法克服了PLS回归易受噪声累积影响的缺点,稳定性和解码正确率均更高,相比传统的降维方法,PLS提取特征个数更少,包含有用信息更多,三组实测数据的解码正确率分别为93.59%、84.00%和83.59%。
Due to the sparse nature of the coding of brain information, the neuron cluster signals recorded by the microelectrode array contain a large amount of noise and redundant information, which reduces the stability and accuracy of motion intent decoding. In order to solve this problem, PLS (Partial Least Squares) feature extraction is applied to the neuronal clustering decoding. First, PLS is used to extract the characteristics of the neuronal cluster front distribution, and then the features are classified by Support Vector Machine (SVM) , Decoded to get the movement intention. The results showed that the decoding method of PLS combined with classification model overcomes the shortcomings that PLS regression is easily affected by noise accumulation and the stability and decoding accuracy are higher. Compared with the traditional dimensionality reduction method, the number of extracted features of PLS is smaller and contains more useful information. The correct rates of the three groups of measured data are 93.59%, 84.00% and 83.59%, respectively.