基于少次刺激脑电的测谎技术研究

The effective of lie detection has been proved by widely practical application. However, the theory of lie detection was not yet perfect, and the research on psychological mechanism was still not deep enough. There were a large number of stimuli in the current Event-related Potentials (ERP)-based lie detection technology.Also,the technology of extracting ERP was still rough, and the accuracy of lie detection was not high enough. We would select P300 to detect liars, which can reflect the nature of cognitive processing for the lying. We propose the acquisition method based on multi-channels EEG in the lie detection research. First,we would analyze the difference from the multi-channels EEG and ERP(P300) evoked by lying and truth-telling. Also, extraction and reconstruct methods of the P300 with high SNR would be researched using the technology of signal processing with multi channels.Furthermore,the extraction technique of some feature parameters that could sensitively reflect the features of the P300 components would be studied. Third, efficient classification model of two kinds of the feature vectors for lying detection would be analyzed and established. Stimuli mode would be also studied here. Finally an effective lie detection method that could be used for practical application would be implemented. Accordingly, the liar could be detected by small-number stimuli using the proposed method. The research and related results would deepen our understanding of the brain's mechanism of processing information during lying, and overcome the disadvantages of current EEG-based detection techniques. The study would enhance the robustness, flexibility and accuracy of lie detection technology, and thus would be used widely in clinic,criminal investigation and anti-terrorism,etc.

测谎的有效性已在诸多实践中得到证实，然而对测谎相关理论及谎言心理机制的研究还不够深入。当前基于脑电事件相关电位（ERP）的测谎技术的刺激量大，ERP的提取技术粗糙，测谎准确率仍较低。本项目选取更能反映谎言的认知加工机理的P300做为测谎工具,首次提出在测谎技术中进行多导脑电的采集，分析人在说谎及说真话时的多通道脑电信号和P300的差异，然后利用多道信号处理技术提取并重建具有高信噪比的P300成分；同时，研究能敏感反映P300成分的特征参量，并研究高效的基于机器学习的谎言分类模型。项目还对测谎刺激模式进行深入研究及改进，以期能实现一种高效且实用性强的测谎方法，该方法基于少次刺激即可自动识别谎言及说谎者。此项目研究内容及成果将深化对谎言过程中人脑信息加工机制的认识， 克服当前基于脑电的测谎技术的诸多弊端，提升测谎技术的鲁棒性、和灵活性和测谎准确率，在临床、刑侦和反恐等方面会有更广应用。

测谎的有效性已在诸多实践中得到证实，然而对测谎相关理论及谎言心理机制的研究还不够深入。 本项目将脑机接口与测谎技术结合，选取更能反映谎言的认知加工机理的事件相关电位—P300做为测谎工具,分析人在说谎及说真话时的多通道脑电信号和P300的变化及差异，然后建立高信噪比的P300成分提取及重建方案；同时，提取能敏感反映P300成分的特征参量，并研究建立谎言识别的分类模型。项目还对测谎刺激模式进行深入研究及改进，以期能实现一种高效且实用性强的测谎方法，该方法最终基于少次刺激即可自动识别谎言及说谎者。在研究过程中，我们对EEG信号的线性、非线性的方法，对时频域的特征提取方法及各种现代模式识别方法进行了广泛研究，并将它们应用到测谎过程中的EEG信号的分析，基于项目的研究内容，严格按照研究计划，我们发表相关研究论文16篇（其中2篇录稿定于2017年8月出版）。这些研究论文涉及到研究内容的各个方面，均涉及到脑电等信号的特征提取和模式识别技术领域的应用研究。此项目研究内容及成果将深化对谎言过程中人脑信息加工机制的认识， 打破当前基于脑电的测谎技术的弊端，提升测谎技术的鲁棒性、灵活性和反测谎水平，在刑侦、反恐等方面会有更广应用。