脑运动神经系统的非线性建模与辨识  

Nonlinear modeling and identification of motor cortical neural system

作　　者：方慧娟[1] 王永骥[1] 何际平[1,2] 刘珊[1] 

机构地区：[1]华中科技大学控制科学与工程系,湖北武汉430074 [2]亚利桑那州立大学生物设计学院,美国亚利桑那州滕比85287

出　　处：《华中科技大学学报（自然科学版）》2009年第1期81-84,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60674105);教育部博士学科点专项基金资助项目(20050487013)

摘　　要：针对大脑运动皮层群体神经元信号与运动轨迹关系的辨识,分别建立了基于人工神经网络(ANN)和基于最小二乘支持向量机(LS-SVM)的非线性具有外部输入的自回归NARX模型.在三维虚拟空间中对猴子手臂运动实验记录的多通道神经元信号进行分析,通过与线性ARX模型的比较,说明非线性模型比线性模型能够更好地描述脑运动神经系统,并且用最小二乘支持向量机建立的模型比人工神经网络建立模型的预测精度更高,泛化能力更强,适用于大脑皮层神经元信号的分析,有利于实现性能更高的脑机接口系统.Nonlinear autoregressive with exogenous （NARX） models based artificial neural networks （ANN） and least squares support vector machines （LS-SVM） were established to identify the relations between the activities of cortical neural ensemble and movement trajectories. The populations of neurons were recorded simultaneously with kinematics of arm movement while the monkey performed center-out task in a three-dimensional virtual environment. The results show that the nonlinear NARX method is better than the linear ARX method to model the cortical neural system. And the LS- SVM based model has higher prediction accuracy and better generalization performance than that of ANN based model. It is seen that the LS-SVM algorithm is suitable for cortical signals analyses and holds hope for a possible more accurate brain-computer interface （BCI）.

关 键 词：脑 计算机接口 算法 神经系统 数学模型 支持向量机 

分 类 号：Q426[生物学—神经生物学] TP391[生物学—生理学]