Cognitive Trace: Neuro-Facial Recognition Analysis

Abstract

Cognitive Trace is a Multimodal Artificial Intelligence system designed to detect cognitive recognition by integrating Electroencephalography (EEG) signals and facial expression analysis. The system primarily focuses on identifying the P300 component in Event-Related Potentials(ERP),which is an involuntary neural response triggered when a subject recognizes familiar stimuli. Alongside this, facial micro-expressions are analysed using a Convolutional Neural Network (CNN) to estimate stress-related emotional states such as fear, anger and surprise. A weighted probabilistic fusion model combines EEG-based recognition probability and facial stress probability to produce a final confidence score. Unlike traditional lie detection system, the proposed system avoids binary classification and instead provides probabilistic outputs, ensuring ethical and non-invasives. Experimental results indicate that the multimodal approach significantly improves reliability, reduce false positives, and enhance robustness against manipulation.

Introduction

The text presents a multimodal AI system designed for recognition detection, aimed at improving reliability in fields such as criminal investigation, security screening, and neuroscience. It addresses the limitations of traditional polygraph-based methods, which rely on indirect physiological signals and are vulnerable to manipulation and inaccuracy.

The proposed system combines two main data sources: EEG brain signals (specifically the P300 Event-Related Potential linked to recognition) and facial expression analysis using computer vision. EEG provides objective neural evidence of recognition, while facial expressions offer complementary behavioral cues such as stress or fear. Together, they form a more robust and less biased detection framework.

The system processes data through a multi-layer architecture: (1) input collection from EEG devices and cameras, (2) preprocessing to remove noise and normalize data, (3) feature extraction to identify P300 signals and facial emotion patterns, (4) a fusion layer that combines both outputs using a weighted probabilistic model (with higher weight given to EEG), and (5) a visualization layer that presents results such as ERP waveforms, emotion classification, and final probability scores in a user-friendly interface.

The system includes three main modules: EEG processing for detecting neural recognition signals, facial expression analysis using CNNs to classify emotions and stress levels, and a fusion module that integrates both outputs into a single probabilistic decision. A visualization module then presents results in real time using interactive dashboards.

Related work shows that traditional polygraph methods are limited, while EEG-based systems (especially P300 “brain fingerprinting”) offer more direct measures of recognition. Facial expression recognition has improved significantly with deep learning, and recent research emphasizes multimodal systems that combine different signals for better accuracy. The proposed system builds on these advances by integrating EEG and facial analysis with a weighted fusion approach.

Conclusion

This paper presents a novel multimodal framework for cognitive recognition detection using EEG and facial expression analysis. The system successfully identifies P300 signals and combines them with facial stress indicators to produce a reliable recognition probability. The proposed approach improves accuracy, ensures ethical usage, and reduces dependency on subjective interpretations. Future work can include real-time deployment, larger datasets, and interpretation of additional modalities such as voice and eye tracking.

References

[1] L.A. Farewell and E. Donchin, “Talking off the top of your head,” Electroencephalography and Clinical Neurophysiology,1991. [2] J. Polich, ”Updating P300: AN integrative theory,” Clinical Neurophysiology,2007. [3] P. Ekman, “Facial Expressions of emotion,” American Psychologist,1993. [4] P. Krizhevsky et al.,”ImageNet classification with deep CNNs,” NIPS,2012. [5] MNE Developers, “MNE-Python Software,” 2020 [6] TensorFlow Team,” TensorFlow: Large-scale ML platform,” 2015. [7] [CK+ Dataset, Carnegie Mellon University of Technology.

Copyright

Copyright © 2026 S. Jasmine, N. Visalini, I. Abinaya, S. Rubinipriya. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.