基于运动想象脑功能网络的节点加权支持向量分类方法研究

项目名称： 基于运动想象脑功能网络的节点加权支持向量分类方法研究

项目编号： No.61201302

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 佘青山

作者单位： 杭州电子科技大学

项目金额： 25万元

中文摘要： 以运动想象脑电信号为对象的脑-机接口不依赖外周神经和肌肉组织的参与，是在大脑与外界之间建立的直接的交流通路，近年来引起国际学术与工程界的积极关注。然而，基于运动想象脑电信号的模式识别研究仍存在一些不足，如：多类分类准确率不高、特征提取及通道选择缺乏自适应性等。现有的研究方法多侧重局部激活脑区的定性与定量分析，忽视了脑区之间的相互影响和协同作用。从脑功能网络角度出发，本项目提出了节点加权支持向量分类方法。主要研究内容为：1）采用多通道脑电信号构建脑功能网络，提出网络整体特征与节点数据局部特征联合表示的特征提取方法；2）研究脑功能网络节点的重要性测度，建立权值函数模型；3）将节点权值作为新的超参数引入最优分类超平面构建中，利用多核学习方法求解节点加权支持向量机具有混合约束的非线性规划问题，通过间接构造法设计多类分类算法；4）通过仿真实验和在线测试进行验证。项目具有重要的理论意义和应用价值。

中文关键词： 节点加权支持向量机；运动想象；脑功能网络；脑电；

英文摘要： Motor imagery EEG-based brain-computer interface has provided a direct communication channel between the brain and environment, which does not depend on the brain's normal output pathways of peripheral nerves and muscles. It has aroused positive concern of the international academic and engineering in recent years. However, the motor imagery EEG-based recognition study still has some disadvantage, such as low multi-class classification accuracy, adaptability deficiency of feature extraction and channel selection. Current methods have emphasized particularly on qualitative and quantitative analysis of locally activated brain regions, which have neglected the interaction and synergy between regions. From the view of functional brain network (FBN), a node weighted support vector classification method is proposed based on FBN in the project.The key includes: 1) Multi-channel EEGs are used to construct FBN, and a feature extraction method is proposed combining global characteristics of FBN and local characteristics of node data; 2) Node weight function is modeled based on its importance measure; 3) Node weights are considered as hyper-parameters into the construction of optimal classification hyperplane, where the nonlinear programming problem with mixed constraints of node weighted support vector machine is solved u

英文关键词： node weighted support vector machine；motor imagery；functional brain network；EEG；