基于PCANet和SVM的谎言测试研究

主成分分析网络（Principal Component Analysis Network，PCANet）是基于深度学习理论的一种非监督式的特征提取方法，它克服了手工提取特征的缺点，目前其有效性仅仅在图像处理领域中得到了验证。本文针对当前谎言测试方法中脑电信号特征提取困难的缺点，首次将PCANet方法应用到一维信号的特征提取领域，并对测谎实验的原始脑电信号提取特征，然后使用支持向量机（Support Vector Machine，SVM）将说谎者和诚实者的两类信号进行分类识别，将实验结果和其它分类器及未使用特征提取的分类效果进行了比较。实验结果显示相对未抽取任何特征的方法，提出的方法PCANet－SVM可以获得更高的训练和测试准确率，表明了PCANet方法对于脑电信号特征提取的有效性，也为基于脑电信号的测谎提供了一种新的途径。

Deception Detection Study Based on PCANet and Support Vector Machine

Principal Components Analysis Network (PCANet)is a feature extraction method based on deep learning theory and unsupervised learning modes,which overcomes the shortcoming of hand-crafted features and its efficiency has been only proved in several literatures for picture processing.In this paper,PCANet is applied to process the one dimensional signals for the first time in order to overcome the disadvantages of hand-crafted features from EEG signals in deception de-tection.PCANet is used to extract features from raw EEG signals in the deception detection experiment.The feature vectors were fed into three classifiers including Support Vector Machine (SVM)to classify the guilty and innocent subjects.The ex-perimental result was compared with the results from other classifiers and the mode of using raw EEG signals as features. The experimental results show that the proposed method PCANet-SVM obtains the highest training and testing accuracy, which indicates the efficiency of extracting features from EEG signals and provides a new solution of detecting lying.