基于深度残差网络的脱机手写汉字识别研究

手写汉字识别是模式识别与机器学习的重要研究方向和应用领域;近年来,随着深度学习理论方法的完善、新技术的层出不穷,深度神经网络在图像识别分类、图像生成等典型应用中取得了突破性的进展,其中,深度残差网络作为最新的研究成果,已成功应用于手写数字识别、图片识别分类等多个领域;将研究深度残差网络在脱机孤立手写汉字识别中的应用方法,通过改进残差学习模块的单元结构,优化深度残差网络性能,同时通过对训练集的预处理,从数据层面实现训练生成模型性能的提升,最后设计实验,验证深度残差网络、End-to-End模式在脱机手写汉字识别中的可行性,分析、总结存在的问题及今后的研究方向。

Recognition of Off-line Handwritten Chinese Character Based on Deep Residual Network

Handwritten Chinese character recognition is an important research direction and application field of pattern recognition and machine learning. In recent years, with the development of the theory and the new technology, deep neural network have made a breakthrough in the field of image recognition and image generation. Specialty, Deep Residual Networks as the latest method, has been successfully applied to handwritten numeral recognition, image recognition classification and other fields. In this paper, we study the application of Deep Residual Networks in off-line isolated handwritten Chinese character recognition, and optimize the performance of Deep Residual Networks by improving the unit structure of residual learning module. At the same time, we improve the model performance by preprocessing the training set. Then, the experiment is designed to verify the feasibility of the Deep Residual Networks and End-to-End mode in off-line handwritten Chinese character recognition. And finally we analyze and summarize the existing problems and future research directions.