Research On Robot Writing Of Chinese Characters Based On Deep Learning

Writing ability is a symbol of human wisdom.In order to construct the self-learning ability of Chinese characters,the research of robot intelligence is a hot topic in recent years.The traditional robotic Chinese character writing technology is limited to the Chinese character structure determined by image processing and data concentration,and the stroke information contained in the Chinese character writing process and the personalized characteristics of Chinese character writing are lost.This thesis combines the deep learning related technology to construct the Chinese character sequence generation model,and uses the Dobot robot and related control algorithms to realize the robot Chinese character writing task,which solves the problem of saving the Chinese character stroke information and diversifying the structural form.The paper mainly studies two methods of robot Chinese character writing: one is based on online handwritten data set,which uses the deletion of redundant points,normalization and other preprocessing operations to reduce the data complexity,and uses long and short time memory networks.The hybrid Gaussian model and the softmax model are used to construct the Chinese character sequence completion model,and the Chinese character handwriting generation model based on deep learning technology is constructed.The other is to complete the trajectory planning of the Chinese character writing task through the robot's forward and inverse kinematics calculation,and realize the docking of the Chinese character generation model and the robot system based on the Dobot robot arm,so that the robot can independently write Chinese characters.Based on the CASIA-OLHWDB online handwritten data set,the paper designs the corresponding visualization software,and compares and analyzes the Chinese character data before and after preprocessing.The preprocessed data set is used to train the Chinese character handwriting generation model to generate various forms of Chinese characters.The DOBOT robotic arm completed secondary development and realized robotic Chinese character writing.The above test results show that the Chinese character writing method based on deep learning allows the robot to learn the style characteristics of a given font library autonomously,and intelligently writes personalized Chinese characters.It provides theoretical and practical reference for the teaching of Chinese writing robots and robots.