Implementation of Machine Learning in BCI Based Lie Detection

In this study, EEG, fNIRS, and HRV signals, recorded from a group of subjects when they were answering a series of true or false questions, were used to see if there is a correlation between BCI results and lying. The EEG and fNIRS signals were collected with g.Nautilus fNIRS-8 headset, while HRV was measured using the Wellue Smart Pulse Oximeter for Adults and Infant connected to iPhone 8 via the ViHealth app. After all the subjects' BCI signals were collected, the raw data was processed in MATLAB and then put in a CSV file. The CSV file was put in MATLAB's Classification Learner KNN and SVM to determine the accuracy of the results. The accuracy of KNN and SVM functions had a range of 75% to 79.4%. The learner was able to predict 81.5% of the truths and 73.7% of the lies.