Individual identification based on facial dynamics during expressions using active-appearance-based Hidden Markov Models

Determining identity of a person is a continually growing subfield of computational intelligence. Measurable biological characteristics, or biometrics, are used to quantify the physical features of an individual for use as a means of identification. There have been psychological studies recently that suggest a new biometric - facial dynamics. In this work, the hypothesis is that facial dynamics of an individual face could be used as an effective biometric for person identification. The method described here applies Stacked Active Shape Models for automated face detection and labeling, Active Appearance Models for feature extraction, and Hidden Markov Models for data analysis. Individual models are constructed for each person in this scenario and used to test identification with new video of facial expressions of the same individuals. Results confirm the hypothesis and demonstrate the efficacy of the potential approach.