Enhanced deception detection through integrated EEG, respiration, and reaction signal analysis using optimized...

Deception, defined as an act that intentionally conceals the truth, has been increasingly studied using neurophysiological and computational approaches. This study evaluates whether combining Electroencephalography (EEG), respiration, and reaction signals can enhance deception detection. A dataset of 30 subjects who participated in the Guilt Knowledge Test (GKT) was analyzed, with subjects divided into guilty and innocent groups. From these signals recorded during the GKT, 71 morphological-temporal, frequency, and wavelet features were extracted. The classification approach utilized the K-Nearest Neighbors (KNN) algorithm and validation methods, with the optimal feature subset determined through a modified Cuckoo Optimization Algorithm (COA). Classification results revealed an accuracy rate of 93.3% using a selected set of 27 features. Additionally, when exposed to specific images, a statistically significant difference (P < 0.05) in reaction time was observed between guilty and innocent subjects. These findings are consistent with, and in some cases comparable to, previous studies that employed psychophysiological signals and EEG, while offering the additional advantage of multimodal integration with feature-level interpretability.