Research And Implementation Of Numbered Musical Notation Identification And Braille Musical Translation Algorithm Based On DSP

As the image processing technique improves, musical notation recognizing technology is becoming a hot research subject. But the current research mainly focuses on the recognizing of stave, while the research on numbered musical notation is less. However, in China, most of these music books are written in numbered musical notations and there are quite a few music books written for the blind.According to the survey of recognizing methods of musical notation by scholars at home and abroad, this thesis designs a system which includes the module of the pre-process of the numbered musical notation image, module of segmentation of notation, module of recognition, module of recombination of the musical notations, module of the translation to the Braille musical notation and the communication between DSP and PC. The pre-process module includes gray-scale transformation、 binaryzation、feature extraction based on the location of bar-line. This system combine methods of Vertical and connected domain labeling to divide the notation on the musical image which has been cleared the lyrics and other words. And, this thesis uses SVM to recognize the musical notations. After that, the thesis reorganizes the notation according to the musical principle. At last, the thesis translates the numbered notation to the blind musical notation for the blind to read. During the process of the DSP hardware design, the thesis firstly transforms the algorithm model based on MATLAB to C language; secondly translate the numbered musical notation to the Braille notation and return the final results to PC using the serial port of DSP. By this way, a complete recognition and translating system is completed. User can read, save and modify the results.This software design is based on Matlab2012、Microsoft Visual Studio2010、 CCS3.3 and hyperterminal, while the hardware is on the base of OMAPL138 board, the key processor is TMS320C674x DSP.This thesis makes use of musical image scanned as RGB type to do tests. According to the experiments, this system can recognize the ordinary numbered musical notation with a high ratio and robustness. This system based on DSP board has realized the fast recognition of numbered musical notation and the correct translation of Braille notation and do some preparation for a future completed system.