基于脑电信号的博弈决策预测方法研究  

Study of prediction method based on decision EEG Game

作　　者：刘新磊[1] 李海峰[1] 马琳[1] LIU Xinlei;LI Haifeng;MA Lin(School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《智能计算机与应用》2018年第5期75-78,82,共5页Intelligent Computer and Applications

基　　金：国家自然科学基金(61171186)

摘　　要：博弈认知状态由于其严密的逻辑性和数学框架,是脑机接口的新兴领域。近年来受到数学家、心理学家和计算机学者广泛的关注,并且已经取得一定的研究成果。首先,本文设计了具有典型博弈活动的"石头-剪刀-布"认知状态研究实验。接着对博弈活动进行了阶段性的划分。然后,采集了17个被试的脑电数据,接着对采集到的脑电数据进行:(1)脑电预览;(2)伪迹剔除与矫正;(3)数字滤波;(4)转参考;(5)Double-ICA去伪迹;(6)脑电分段;(7)基线矫正等离线预处理操作。在此基础上,分别从统计学特征、时域特征和频域特征3个方面对被试出石头、剪刀、布3种情况下的脑电信号进行特征提取,使用基于关联的特征选择方法进行特征选择。最后使用SVM分类器进行分类识别,其中SVM的核函数采用线性核函数,达到了83.3%的识别率。Game cognitive state,due to its rigorous logic and mathematical framework,is an emerging field of brain-computer interface. In recent years,it has drawn much attention from mathematicians,psychologists and computer scholars,and has achieved some research results. At first,this paper designs a cognitive state research experiment ＂stone-scissors-cloth＂,which is a typical game. Secondly,the paper divides the game into different stages. Then,the paper collects 17 subjects of EEG data,and with these EEG data,respectively performs the following steps：（ 1） electrical preview;（ 2） artifact correction;（ 3） filter;（ 4） reference;（ 5）Double-ICA artifact correction;（ 6） epoch;（ 7） baseline correct offline preprocessing operations. On this basis,the paper extracts the features of EEG from stone,scissor and cloth in terms of statistics,time domain and frequency domain; next,selects features by using feature selection method based on correlation. Finally,the research uses the SVMclassifier for classification recognition,with linear kernel function as the SVMkernel function,and the recognition rate reaches 83.3%.

关 键 词：博弈 统计特征 功率谱 留一法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]