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

Area of Science:

• Neuroscience
• Computational Psychology
• Forensic Science

Background:

• Deception detection is increasingly studied using neurophysiological and computational methods.
• Existing research often focuses on single signal types, limiting comprehensive analysis.
• Multimodal approaches integrating various physiological signals show promise for enhanced accuracy.

Purpose of the Study:

• To evaluate the enhancement of deception detection by combining Electroencephalography (EEG), respiration, and reaction signals.
• To determine the optimal feature subset and classification algorithm for multimodal deception detection.
• To assess the statistical significance of physiological differences between guilty and innocent subjects.

Main Methods:

• Analysis of a dataset of 30 subjects undergoing the Guilt Knowledge Test (GKT).
• Extraction of 71 morphological-temporal, frequency, and wavelet features from recorded EEG, respiration, and reaction signals.
• Classification using K-Nearest Neighbors (KNN) algorithm, optimized with a modified Cuckoo Optimization Algorithm (COA) for feature selection.

Main Results:

• Achieved a classification accuracy of 93.3% using a selected subset of 27 features.
• Identified a statistically significant difference (P<0.05) in reaction time between guilty and innocent subjects when exposed to specific images.
• Demonstrated the effectiveness of the multimodal approach within an interpretable, data-efficient computational pipeline.

Conclusions:

• Combining EEG, respiration, and reaction signals provides a robust enhancement for deception detection.
• The developed computational pipeline offers a structured and efficient method for analyzing multimodal psychophysiological data.
• Findings support the utility of multimodal physiological signal integration in controlled experimental settings for deception detection.