Deep Learning Applications in Brain Computer Interface Based Lie Detection
Mohammad Affan Khalil, Johnny Can, Kiran George
Abstract
Improvement of Brain Computer Interface (BCI) based-lie detection was investigated using the bio signals collected from five subjects and processing with various deep learning algorithms. The bio signals collected from each subject were Electroencephalography (EEG), Functional near-infrared spectroscopy (fNIRS), Heart Rate Variability (HRV), and Blood Oxygen Saturation. The EEG and fNIRS signals were gathered using g.Nautilus fNIRS-8 headset while the HRV and Blood Oxygen Saturation data were collected through Wellue Smart Pulse Oximeter for Adults and Infant, and stored on an iPhone X using ViHealth app via Bluetooth. After pre-processing each of subjects' BCI signals, they were stored in a CSV file and later processed in MATLAB to compare performance of various deep learning models. The accuracies of the models ranged from 44.0% to 86.0%. The deep learning model with the 86.0% accuracy predicted 97.3% of the truths and 53.8% of the lies.