Similar Pattern Recognition Based on Hopfield Neural Network Associative Memory

The associative memory function of discrete Hopfield neural network is widely used in the field of pattern recognition because of its great fault tolerance. In this paper, for the crosstalk problem between similar memory samples in the associative memory of discrete Hopfield neural network, an improved Hopfield neural network associative memory pattern recognition algorithm is proposed based on neuron excitation threshold adjustment. Besides, the fault tolerance and real-time performance of the proposed algorithm are verified through the recognition of similar speed limit traffic sign images. The simulation results show that the correct recognition rate can still reach more than 90% when the pattern to be recognized is contaminated by noise to the extent of 50%. It follows the simulation results that it has the ability to recognize incomplete input patterns and has good real-time performance. Therefore, the improved algorithm proposed in this paper can effectively recognize similar memory samples in the process of associative memory.