The Research On Braile Music Recognition And Segmentation

Purpose——Driven by the "Internet +",the demand for the work of the blind groups has gained more and more attention and recognition,and the blind musicians have increased their chances of creating and communicating in braille,but it is still dependent on those who know the rules of braille and the clear music theory to communicate with people with normal vision,resulting in the time,content and quantity are difficult to meet the growing demand for braille music exchange.The research on the recognition and segmentation of braille music is an effective means to solve the above problems.Methodology——Through the analysis of the composition structure of braille music,in order to overcome the shortcomings of traditional methods in solving the problem of recognition and segmentation,using deep learning framework to design the model,through a large number of sample data training,let the computer automatically learn its feature attributes,at the same time after repeated iterative training to make it have stronger robustness and generalization ability,form a braille music recognition and segmentation method based on deep learning model.Findings——Firstly,a model of braille music image recognition based on convolution neural network(CNN)is designed.The preprocessing image data are sent to the designed model for training,testing and tuning,and finally the method of recognition of braille music images based on CNN achieves a good accuracy.Secondly,in order to solve the problem of lack of braille music corpus,the braille music corpus was created manually,the braille music symbol segmentation model based on long and short term memory network(LSTM)was designed,the model was trained and optimized,and a good segmentation effect was obtained.Research limitations——The emphasis on theoretical research and small data set experiment needs some follow-up work:(1)Considering from the deep learning model,increasing the size and size of the data set in order to enhance the training effect of the model;(2)In practical terms,it is necessary to combine the recognition and segmentation of braille music images,and further expand to the research of part-of-speech tagging,music notation conversion and music melody similarity recommendation.Practical implications——Through the recognition of braille music and segmentation,the recognition of braille music symbol in braille music image realized,and it can also solve the problem of segmentation before the conversion of braille music.Value——Using CNN model to identify braille music image to obtain corresponding braille music notation;combining with the created braille music corpus,the LSTM model is used to segment the braille music,and the result of segmentation is obtained,which is convenient for further processing in subsequent research.