| Lying is prevalent in all aspects of society.Exploring the cognitive mechanism of lying using electroencephalography(EEG)is one of the most exciting areas of cognitive neuroscience research.Previous studies using event-related potentials(ERP)and functional connectivity network theory have looked at the complex information exchange patterns of the human brain while lying from a variety of angles,but research using the sliding window method has a number of drawbacks,including the difficulty of determining the window scale,the complexity of statistical analysis,and the loss of data resolution.In this study,Hidden Markov Model(HMM)was used to replace the sliding window method,and the existence time period of each cognitive state in the lying process was self-divided from a data-driven perspective.The s LORETA method was used as the source localization method to investigate the activated brain regions during the main state of the lying process.For this main state,the PLI index was used to assess the functional connectivity strength of each brain region.Finally,a classification model was created using the SVM classifier.This research will help researchers better understand how lies are processed in the brain,as well as improve the robustness,flexibility,and accuracy of lie detection technology:This study recruited 30 normal subjects and divided them into two groups based on the principle of equality: the guilty group and the innocent group.The experimental data of the two groups of subjects was then collected using the tri-stimulus paradigm.First,the ERP waveforms of the whole brain electrodes were analyzed,and the Fz electrodes were chosen to investigate the P300 component’s latency and amplitude.The brain state changes of the two groups of subjects were then analysed and the state characteristics were statistically analysed using the hidden Markov model.After determining the main state,the cerebral cortical activation areas of the two groups of subjects were analysed using Standardized low-resolution brain electromagnetic tomography(s LORETA)during the time period of the state.The Phase Lag Index is then used to calculate the functional connectivity matrix of this principal state(PLI).Finally,an efficient polygraph system is built by combining the state’s temporal distribution features and using the SVM classifier for 6-fold cross-validation.Through the above research,this paper finds that: ERP analysis results showed that the latency and amplitude of P300 in the guilty group were significantly higher than those in the innocent group,indicating that lying requires complex cognitive inhibition and working memory processes;the results of the hidden Markov model showed that the guilty group compared with the innocent group.In the innocent group,the dynamic transition of brain states was more consistent,and the evaluation indicators such as state occupancy rate and duration in the P300 distribution time period were significantly different,indicating that this time period reflects the unique cognitive process of guilty.The source tracing results showed that the activation of the occipital lobe,prefrontal lobe and bilateral temporal lobes of the guilty group was much stronger than that of the innocent group during this state period,indicating that the guilty group performed complex cognitive functions by invoking the appeal brain area during this time period.The results of the PLI connection matrix showed that the phase synchronization between the FCz-FC3,FCz-O2 and other electrode pairs in the guilty group was significantly stronger than that in the innocent group,indicating that the information exchange between the appellate brain areas plays a role in the cognitive processing of lies.The classification results show that after the feature vector fused with state time information is trained by SVM,the highest classification accuracy of the test set is 95.12%,indicating that the lie detector system can be used in the field of lie recognition. |
PDF Full Text Request