Research On Polygraph Algorithm Based On P300

With the constant improvement of sociocultural awareness and levels of technology, it’s a major problem needed to solve for public security to study how to use the technology means in criminal investigation. As a high-tech means, polygraph technology determines whether the subjects have lying behavior by monitoring physiological changes when they answer questions, and not test the lie itself, thus eliminating the innocent and refine search, and then achieve a breakthrough for the further investigation of the case. At present, many developed countries have talked about polygraph technology as an adjunct of criminal investigation, and it has proved that polygraph technology popularization and application in the trial work is one of the effective means and ways to improve the work efficiency.In recent years, thought P300polygraph technique is favored by more and more researchers, which is based on psychology and information science theoretical, and they have made great progress, but the current research is still unable to meet the real needs, which is because the usual experimental program designed from the real crime scene, and classification used commonly is not precise enough, and it’s difficult to accurately identify criminals and innocent. For these reasons, this study extracts EEG of15volunteers though design a crime experiment, and that compare with the biographical information to analysis the similarities and differences of ERP waveform between different experiments, and then combine with new classifier to explore the key that improve the accuracy of polygraph. The main work is as follows:First, for the current experiments there are some issues that ignoring the effects on P300from suspects’ mental, and programs designed are divorced from the real case and it’s unable to assess whether subjects coordinate in programs, and this paper presents that it can join trial session before ERP test. Though crime scene reconstruction ways to enhance the criminal psychology and deception awareness of the subjects, it can make the crime scenes are more realistic simulation. Meanwhile, using random questions are asked to ensure the subjects focused in the ERP testing process in both experimental groups, and as a standard test to assess the degree of match, thus ensuring scientific rationality collected data. It’s found that biographical information of subjects is likely to inducing a strong P300by analyzing the two sets of experiments, and it illustrates that enhancing the familiarity of information is the critical of evoked P300significant in the design of polygraph test.Then, considering the low classification accuracy of SVM used in polygraph and strong individual differences maybe exist in EEG data of single-channel, based on the analysis of relationship between K-nearest neighbor algorithm and SVM, there propose combining the two algorithms to form a new classifier K-SVM which is applied to the polygraph research. It’s found that recognition rate is improved to some extent in two groups by using K-SVM to classify the experimental data of two groups respectively, and it provides some reference value for polygraph study.Finally, given the kernel function and parameters of SVM algorithm is difficult to determine and the larger impact on classification results, we propose to use cuckoo algorithm which has fewer parameters and simpler algorithm and global search ability to optimize the parameters of SVM, and constitute CS-SVM classifier, and then using the best parameters of find a classification of the data and found that there is a high classification rate by using CS-SVM than mere SVM, and it proofs that CS-SVM can be able to better used in the classification polygraph test.