Investigation of Human Brain Waves (EEG) to Recognize Familiar and Unfamiliar Objects Based on Power Spectral Density Features

Abstract

Research into the application of EEG technology for lie detection during interrogation has gained significant popularity. However, no EEG method has yet proven to be entirely reliable for lie detection. Therefore, further research is necessary to develop a roadmap for utilizing brain signals in interrogation tools other than the Polygraph, which is still commonly used by law enforcement to solve crimes. This additional research is expected to yield valid data and more dependable methods for analyzing EEG signals. The parameters obtained from this research can be used to develop AI-powered computer systems that can detect when someone is lying based on their brain signals. This study used Power Spectral Density (PSD) analysis to investigate brain activity in 20 participants who viewed familiar and unfamiliar images. EEG data were collected from specific channels (T3, T4, T5, T6) in the temporal region, as well as channels (O1, O2) in the occipital region, across the alpha, beta, and gamma frequency ranges. The findings revealed that the PSD values observed on the specified channels T3, T4, T5, and T6 were higher when participants did not recognize the image object. Additionally, channel O2 showed increased right-brain activity when participants failed to recognize the object. Machine learning algorithms were employed to classify the data, with the Random Forest method achieving the highest accuracy at 95.4%.