In this thesis, a real-time recognizing system is studied for the printed piano score. The proposed system is composed of three parts. First, the vision system, which is equipped on a robot hand, captures and recognizes a part of the whole musical score image. Then, the recognized musical data are transferred to the sound synthesis system to play the music. The robot is employed to overcome the resolution problem by capturing the piano score at a closer distance. The recognition system operates in real time so that, once the recognizing operation starts, the computer plays the music without stop.
The above system is implemented on a 32-bit personal computer along with the corresponding musical score being played by a speaker built in the personal computer. Experiments show that it takes 9-13 seconds to start playing the piano by synthesis and 20-25 seconds to recognize a sheet music by capturing twice the sheet music and the recognition success rate is about 96 percent.