Abstract:
This paper presents an overview of a number of classical face recognition techniques, including an assessment of their suitability for applications in real-time embedded systems such as digital cameras, and a generalized discussion of the training requirements for different face recognition techniques. Also, an introduction to face recognition approaches based on hidden Markov models (HMM) is made, and their advantages and benefits are shown, particularly their ability to provide acceptable levels of accurate recognition for unseen collections of facial images. Finally, a detailed description/implementation of a working HMM test bed is done, together with a presentation of some preliminary results.
