基于实时样本采集的个性化手写汉字输入系统设计

Design of Handwritten Chinese Character Input System Based on Real-time Sample Collection

手写汉字识别是手写汉字输入的基础;目前智能设备中的手写汉字输入法无法根据用户的汉字书写习惯,动态调整识别模型以提升手写汉字的正确识别率;通过对最新深度学习算法及训练模型的研究,提出了一种基于用户手写汉字样本实时采集的个性化手写汉字输入系统的设计方法;该方法将采集用户的手写汉字作为增量样本,通过对服务器端训练生成的手写汉字识别模型的再次训练,使识别模型能够更好地适应该用户的书写习惯,提升手写汉字输入系统的识别率;最后,在该理论方法的基础上,结合新设计的深度残差网络,进行了手写汉字识别的对比实验;实验结果显示,通过引入实时采集样本的再次训练,手写汉字识别模型的识别率有较大幅度的提升,能够更有效的满足用户在智能设备端对手写汉字输入系统的使用需求。

Handwritten Chinese character recognition is the basis of handwritten Chinese characters input system.At present,Handwritten Chinese character input method in smart devices can not adjust the recognition model dynamically according to the user's Chinese writing habits in order to improving the correct recognition rate of handwritten Chinese characters.Through the study of the latest deep learning algorithm and training model,a design method of handwritten Chinese character input system based on the real-time acquisition of handwritten Chinese characters samples is proposed.This method will capture the user's handwriting Chinese characters as the increment of sample.Through re-training the pre-training model,the recognition model can better fit the user's writing habits,and enhance the recognition rate of the Chinese characters handwritten input system.Finally,on the basis of the theoretical method and combining the new design of depth residual network,a comparative experiment of handwritten Chinese character recognition is carried out.The experimental results show that by introducing real-time samples to training,the rate of recognition of handwritten Chinese characters recognition model is improved.It can more effectively satisfy the user's demand for handwriting Chinese character input system at the intelligent terminal.