Research Projects

Deepfake detection has become a critical area of research in the forensic community, with most methods relying on inconsistencies in biometric patterns like eye blinking rate, eye gaze direction, mouth movement patterns, ear placement, and iris color. However, these biometric features are prone to variations from person to person due to differences in social background, culture, behavior, and ethnicity. This research aims to find robust identification parameters of human beings that can be detected from human facial videos. The primary candidate for investigation will be human physiological parameters, specifically heart rate variability (HRV) and its correlation with affectional states of humans. The research will build a human affective recognition based model for deepfake detection, which will perform affective recognition by extracting physiological parameters from spatial domain facial video analysis and computations. The proposed scheme will be validated against gold standards using Photoplethysmography (PPG) sensors, as the current state-of-the-art has proven PPG sensors to produce gold standard values of heart rate. This approach will help identify synthetic human facial videos based on affection recognition, assuming that emotions found in humans are missing in synthetically generated videos.