利用小波分解和支持向量机的心理意识真实性识别研究

采用小波分解和支持向量机(SVM)技术,提出了一种对说谎脑电(EEG)信号特征进行分类的方法,将其应用于心理意识真实性的检测,获得了满意的结果.以真伪已明确的有意义的个人信息(如姓名、生日)作为被测试的隐藏信息,应用隐藏信息(CIT)测试模式对15名受试者各进行两组测试,并记录其脑电(EEG)信号.提取了探测刺激和无关刺激诱发EEG信号的小波系数,并应用具有统计学意义的特征参数作为SVM分类器的输入进行识别分类.实验结果显示,应用leave-one-out交叉验证法对30组样本数据进行训练测试,获得平均正确识别率为88.3%.因此,该方法可以作为一种心理意识真实性检测的新方法,具有无创、较高正确检测率等优点.

Identification of Mentality Facticity Based on Wavelet Decomposition and Support Vector Machines

A practical strategy for classifying the lying electroencephalograph (EEG) characters by wavelet decomposition and support vector machines (SVM) techniques is presented to get a satisfactory results in identifying the mentality facticity. Some significant personal information is ensured, such as name and birthday, and selected as the concealed information. 15 subjects participate in two groups of concealed information tests (CIT) and their EEGs are recorded. Applying wavelet decomposition, the wavelet coefficients corresponding to EEG evoked by probe information and by irrelevant information respectively are evaluated. Then the feature coefficients containing statistical significance are extracted as the input parameters of SVM. 30 samples are chosen to train and test the performance of classifier by leave-one-out cross-validation, 88. 3% accuracy can be achieved in probe information detection.