Digital Recognition Based on Combined Feature and PSO-BP Algorithm

A new method of digital recognition based on combined features and PSO-BP(particle swarm optimization - backpropagation) algorithm is presented.The combined feature vectors of digital image are composed of grid feature,projection feature and structural features of Euler number in accordance with the feature weight coefficients.By applying PSO-BP neural network to recognition,it gives full play to global optimization capability of the particle swarm algorithm and local search advantages of BP algorithm.Experimental results show that this method is of high recognition rate,fast network convergence speed and high precision.