A Face Recognition System Using Neural Networks with Incremental Learning Ability
Soon Lee Toh, Seiichi Ozawa · 2003
This paper presents a fully automated face recognition system with incremental learning ability that has the following two desirable features: one-pass incremental learning and automatic generation of training data. As a classifier of face images, an evolving type of neural network called Resource Allocating Network with Long-Term Memory (RAN-LTM) is adopted here. This model enables us to realize efficient incremental learning without suffering from serious forgetting. In the face detection procedure, face localization is conducted based on the information of skin color and edges at first. Then, facial features are searched for within the localized regions using a Resource Allocation Network, and the selected features are used for in the construction of face candidates. After the face detection, the face candidates are classified using RAN-LTM. The incremental learning routine is applied to only misclassified data that are collected automatically in the recognition phase. Experimental results show that the recognition accuracy improves without increasing the false-positive rate even if the incremental learning proceeds. This fact suggests that incremental learning is a useful approach to face recognition tasks.