Handwriting Posture Prediction Based On Unsupervised Model

Writing is an important basic skill for humans.To acquire such a skill,pupils often have to practice writing for several hours each day.However,different pupils usually possess distinct writing postures.Bad postures not only affect the speed and quality of writing,but also severely harm the healthy development of pupils' spine and eyesight.Therefore,it is of key importance to identify or predict pupils' writing postures and accordingly correct bad ones.The existing method is to find the posture problems of students by taking pictures.The disadvantage of this method is that it is too complicated and the workload is huge.In this paper,we present for the first time a written pose prediction based on an unsupervised model.Pupils' handwriting is used to predict their writing posture problems.First,we propose a pose classification based on expert experience.The atomic errors of the writing posture are classified,the handwritten appearance is summarized and analyzed,and then the expert's experience is combined to build a many-to-many relationship.Secondly,we use handwriting recognition based on multi-task learning.Training the public handwritten data set,through the construction of multi-task learning model,and then through the loss function integration optimization and learning rate adaptive optimization of the two methods,compared to single-task learning,we built the model to improve the accuracy of handwriting recognition;Finally,We propose a pose prediction based on an unsupervised model.The transfer learning combine the neural network of small convolution kernel structure with the multi-task model of handwriting recognition,constructs a handwritten fixed feature extractor,extracts features from pupils' handwritten data,and then reduces feature congestion and combines unsupervised learning and A data model analysis of the written pose determines the pose problem.The experimental data show that after the model fusion,with the improved accuracy of the handwriting,the fixed feature extractor also improves the feature extraction effect of the handwriting.The multi-task handwriting recognition model proposed in this paper is compared with the unfused model.And the fusion single task recognition handwriting model has a certain improvement in the accuracy of the final predicted posture problem.Therefore,the fusion scheme of the multi-task handwriting recognition model is adopted,so that the accuracy of the handwritten prediction posture problem can reach 93.3%,which is significantly higher than the 76.67% accuracy of human experts.