Selection and Efficient Use of Local Features for Face and Facial Expression Recognition in a Cortical Architecture
Masakazu Matsugu · 2007
In this chapter, we reviewed our previously proposed leaning methods (unsupervised and supervised) for appropriate and shared (economical) local feature selection and extraction for generic face related recognition. In particular, we demonstrated feasibility of our hierarchical, component based visual pattern recognition model, MCoNN, as an implicit constellation model in terms of convolutional operation of local feature, providing a substrate for generic object detection/recognition. Detailed simulation study showed that we can realize face recognition as well as facial expression recognition efficiently and economically with satisfactory performances by using the same set of local features extracted from the MCoNN for face detection.