The Research On Method Of Handwritten Character Recognition Based On Extremelearning Machine

With the development of science and technology,a large number of handwritten character information was produced by people.Taking into account the security and privacy of the information expressed by these characters,it is imperative to make the machine to realize the handwritten characters fast and accurately.The mainly method used by handwritten character recognition is optical character recognition.But because of its low recognition rate and high cost,it has not been widely used.At present,template matching,neural networks and support vector machine has been put into the recognition of handwritten characters.In this paper,based on the poor real-time performance and high cost of the traditional character recognition method,propose to use the extreme learning machine algorithm to realize the recognition of handwritten character.Firstly,this paper introduces the definition,basic system and basic method of pattern recognition.The method of pattern recognition based on neural network is introduced,and the working principle and characteristics of the neural network are analyzed and studied.Then propose to use the extreme learning machine to realize handwritten character recognition,according to the unbalance between the structure risk and experience risk of original extreme learning,put forward to use R-ELM and DFT-ELM to realize the handwritten character.A handwritten character recognition system based on extreme learning machine is designed in this paper.The system uses four algorithms include BP neural network,extreme learning machine,regular extreme learning machine and DFT-ELM to realize the recognition of handwritten character.Specific steps including preprocessing,feature selection and dimension reduction and other specific process.The training samples of handwritten character recognition algorithm is the 10000 digital samples from the MINIST sample library,the number of test samples is 1000.Compared the handwritten character recognition results between BP neural network,extreme learning machine,R-ELM and DET-ELM four algorithms.Except this,a simulation was designed to analyze the impact of the number of neurons in the hidden layer.Through the comparative analysis of the simulation results of the four algorithms,as the most classical neural network algorithm,BP achieved a high level of accuracy in the recognition of handwritten numeral recognition results.The extreme learning machine algorithm has a great advantage over the training time of BP neural network.But the recognition accuracy is lower than that of BP neural network.Two optimization algorithms based on extreme learning machine,that is,R-ELM and DFT-ELM,compared with the original extreme learning machine,the generalization ability of the algorithm was improved,and the recognition accuracy of the handwritten numeral characters was improved.