Face Tracking and Recognition in Videos : HMM Vs KNN

The problem of tracking and recognizing faces in real-world, noisy videos is addressed. While traditional face recognition is typically based on still images, face recognition from video sequences has become popular recently. This work describes a new method to perform face recognition from video sequences. Faces are detected, tracked and recognized in a video sequence using Hidden Markov Model and K-Nearest Neighbor. Feature extraction for tracking and recognition is performed by Principal Component Analysis. This process also allows locating and extracting facial feature regions around the eyes, nose and mouth. Identity of the tracked subject is established by fusing pose-discriminant and person-discriminant features over the duration of a video sequence. This leads to a robust video-based face recognizer with the state-of-the-art recognition performance. The quality of tracking and face recognition on real-world noisy videos as well as the standard Honda/UCSD database is tested