A massive parallel neuromorphic computing model for intelligent text recognition

Summary form only given. In this talk, we present a parallel neuromorphic computing model and its application in context-aware Intelligent Text Recognition (ITR). The information processing flow of the proposed computing model imitates the function of a neocortex system. It incorporates large number of simple and fuzzy pattern detection modules with advanced information association layer to achieve perception and recognition. The ITR system based on this computing model serves as the physical layer of machine reading. The system learns from what has been read and, based on the obtained knowledge, it forms anticipations of the word and sentence level context, which helps image recognition. The proposed neuromorphic computing model is naturally massive parallel, hence ideal to be implemented on future many-core processors. Experiments show that the proposed computing model provides robust performance in recognizing images with large noise.