| There is a long history of study on lie-detection.The techniques of lie-detection have also undergone a process from simple to complex.With the development of psychology,brain science and other disciplines,lie-detection technology will be with multiple index.This paper realized the lie-detection by combining the electroencephalogram(EEG)signals with functional near infrared spectroscopy(fNIRS)signals obtained from simulated crime scenes,and the main research work is as follows:1)A simulated crime experiment based on hidden information testing was designed and carried out.In this experiment,EEG signals and functional near-infrared spectroscopy signals of brain were collected synchronously.By using the difference of subjects' response under different stimuli in simulated crime experiment,the feasibility of lie-detection based on fNIRS signals and event-related potentials(ERP)that is a component of EEG was proved.2)The EEG signals and fNIRS signals were combined to realize the lie-detection.The wavelet packet energy entropy of ERPs,the linear regression coefficients of oxyhemoglobin content and cross approximate entropy of fNIRS signals were extracted as the characteristics of EEG and fNIRS signals respectively.The above features were fused.3)An optimization algorithm based on artificial fish swarm algorithm was proposed to optimize the number of nodes in hidden layer of extreme learning machine(ELM),so as to improve the recognition rate.The optimized classifier was used to classify features of single signal and multi-modal signal respectively,and the effectiveness of the optimization method is proved by comparing the recognition rate with that of the traditional extreme learning machine.The results show that the recognition rate of lie-detection method combined with EEG signals and functional near-infrared spectroscopy signals is higher than that of lie-detection method based on single signal,reflecting the effect of multi-modal method of lie-detection. |
PDF Full Text Request