基于cf-svm的眼动信号检测算法.docVIP

基于曲线拟合和SVM的眼动信号分类算法研究* 侯传宇1,2，吕 钊2，吴小培2 （1.宿州学院信息工程学院,安徽 宿州 234000；2.安徽大学 计算智能与信号处理教育部重点实验室, 安徽 合肥 230039) 摘要EOG的眼动信号分类算法精度，改进基于EOG的人-机交互系统性能，论文提出了一种基于曲线拟合(curve fitting)与支持向量机（SVM）的眼动信号分类算法(CF-SVM)，并设计了新的实验范式，增加了“扫视保持”环节。该算法采用曲线拟合方法进行特征提取，在此基础上，使用SVM分类器对眼动信号进行分类。实验室环境下，对9名眼部活动正常的受试者进行了眼动数据采集与识别，CF-SVM算法的平均分类准确率达到98.3%，与其它几种眼动识别方法相比较，其平均正确率分别提升了9.4%、5.9%、1.0%。实验结果表明，CF-SVM算法在眼动信号识别中表现了良好的性能，具有高的分类精度和鲁棒性。 关键词 TN911.7;TP399 　　　 Research on a classification algorithm of eye movements signals based on curve fitting and support vector machine HOU Chuan-yu1,2，LV Zhao2，WU Xiao-pei2 (1.School of Information Engineering, Suzhou University, Suzhou 234000；2.The Key Laboratory of Intelligent Computing Signal Processing, Anhui University, Hefei 230039, China) Abstract：To improve the classification accuracy of eye movement signals and the performance of the BCI system based on EOG signals, this paper proposed a classification algorithm of eye movement signals based on curve fitting and support vector machine (CF-SVM)，and increased a new experimental paradigm named “saccades keeping”.It applied a curve fitting method to extract feature parameters of eye movement signals, on this basis, classified the EOG signals using SVM classifier. In laboratory environment, the experimental data contained eye movements of 9 normal subjects, and the average classification accuracy of CF - SVM algorithm achieves 98.3%, increase 9.4%, 5.9% and 1.0% respectively compared with other identification methods. Experimental results reveal that the CF-SVM algorithm has high classification accuracy and good robustness in EOG signal identification. Key words: human-computer interface; electrooculogram; experimental paradigm；curve fitting; support?vector?machine 0引言 近年来，随着微电子和计算机技术的快速发展，基于眼动信息的人机交互(human-computer interface，HCI)技术已经成为信息技术领域的热点，它的发展为了特殊群体(视觉或运动功能受损的病患或老年群体)和特殊环境下的人-机通信提供了一种新的实现模式[1-3]。同时，基于眼动信息的行为识别研究也成为了模式识别领域的一个新方向，相关研究具有重要的实际意义和学术价值。 眼动信号的采集方法有搜索线圈法、红外线眼动图法、视频记录法、眼电图法(Electrooc