Lie Detection using Facial Expressions and Voice Recognition

Lie detection is a critical area of study, especially within security and forensic applications. This project presents a multimodal lie detection system that analyzes facial micro expressions and vocal patterns to improve the accuracy of deception detection. The system leverages Long Short-Term Memory (LSTM) networks to process sequential data from facial expressions, detected via OpenFace, and audio features extracted through Librosa. Integrating these modalities allows the model to detect subtle cues in both the face and the voice, achieving an accuracy of 85 percent. By employing deep learning on multimodal data, this project advances traditional methods of lie detection by offering a noninvasive, automated approach. The system’s high accuracy and real-time performance hold potential for real-world applications in interview analysis, security screening, and mental health diagnostics.