The Design And Implementation Of Handwritten Digit Recognition System Based On Spiking Neural Network

Artificial intelligence technology has achieved impressive results and tremendous progress,and it is generally believed that there is a huge technological gap from the ultimate realization of the pursuit of computer scientists.Brain-inspired intelligence is considered to be a possible technical approach to general artificial intelligence,and it is also one of the research focuses of the new generation of neural artificial intelligence.Spiking Neural network is a key part of brain-inspired intelligence,and it is also the most biologically meaningful network model at present.In this thesis,the digital recognition of the main digital network of the basic method is researched and the digital recognition system is designed and implemented.The work of the thesis is as follows:(1)The CSNN model is constructed by combining the feedforward SNN model with the convolution,and the recognition rate on the test set reached 99.2% when trained with MNIST.(2)Train the model on the classic MNIST dataset and the neuromorphic dataset N-MNIST,and explore the difference in the performance of SNN models when trained with static and dynamic datasets.(3)A handwritten digit recognition system is built based on the CSNN model,in which the handwritten digit collection end is an Android system tablet,and the digital inference end is an ARM-based Raspberry Pi board.In this thesis,the spiking neural network is successfully applied to the task of handwritten digit recognition,which has good theoretical significance and engineering application prospects.