Facial Image Recognition Based on Connectionist Model and Coarse-to-Fine Pyramidal Decision Strategy

Abstract

In this paper, we propose a novel approach for facial image recognition. Coarse and fine resolutions respectively provide the global and locally detailed information；and we use on SOFM for automatically learning facial features for one resolution. A particular classification strategy, called the coarse-to fine pyramidal decision strategy, is proposed to hierarchically integrate the recognition decision of each resolution with the order from the coarsest resolution to the finest. Potential candidates are sifted as the resolution increases, and fewer and fewer candidates are survived. The experimental results show 100% accuracy is achieved.