A new approach for concealed information identification based on ERP assessment

Area of Science:

• Neuroscience
• Machine Learning
• Forensic Science

Background:

• Concealed Information Test (CIT) studies frequently utilize event-related potentials (ERPs) due to their proven validity in practical applications.
• Existing methods for analyzing ERPs in CIT can be complex and computationally intensive.

Purpose of the Study:

• To propose and evaluate a new approach for identifying concealed information using wavelet coefficients (WCs) and kernel learning algorithms.
• To assess the effectiveness of combining kernel principal component analysis (KPCA) with a support vector machines (SVM) classifier for concealed information detection.

Main Methods:

• 16 subjects participated in a designed CIT paradigm, with multichannel electroencephalogram (EEG) signals recorded.
• High-dimensional WCs of ERPs were extracted across delta, theta, alpha, and beta frequency bands.
• Kernel Principal Component Analysis (KPCA) was used for dimensionality reduction, followed by Support Vector Machines (SVM) classification.

Main Results:

• Significant differences (P < 0.05) were found in WCs features between concealed and irrelevant information.
• KPCA effectively reduced feature dimensionality while enhancing the generalization performance of the SVM classifier.
• The combined KPCA-SVM approach achieved a high accuracy of 93.6% in distinguishing concealed from irrelevant information.

Conclusions:

• Wavelet coefficients offer significant discriminative features for concealed information detection within ERPs.
• The integration of KPCA and SVM provides an effective and accurate method for identifying concealed information.
• This novel approach presents a valuable tool for practical concealed information detection applications.