A character recognition system based upon adaptive feature-detecting and a new model of holographic memory

A novel algorithm based on holographic memory and adaptive feature detection is proposed for pattern recognition. The memory is organized through unsupervised learning. In this model once the training patterns are all fed to the network, the memory will be stable although it may be modified in some cases. Since the adaptive feedback interferes directly only with the feature mapping, the memory system will avoid the risk of being trapped in an unstable state. As a simulation the authors constructed a handwritten Chinese character recognition system with this model