Research And Realization On The Key Technology Of The Musical Note Recognition And Tracking

In Digital Era, plenty of digital music enrich people's lives and bring great challanges. More and more people preferred to get and learn these music more easily. Therefore, the technology of query-by-humming and music automatic transcription are developing quickly. All of these technology are based on musical note recognition. Under the background, firstly, the paper introduces basic music theory and methods of anaylzing audio singal in frequency domain and time domain.Secondly, algorithm of extracting pitch and time length of the musical note are researched, the matching algorithm is analyzed. At last, the system of musical note recognition and tracking is designed, and the result of test is given.For extracting pitch, the dissertation compares auto-correlation function(ACF), Subharmonic-to-Harmonic Ratio(SHR) and Praat(ac) algorithms. And then, the simulation test for sine singal, single musical note, and musical fragment is given.For detecting the duration of the music note, a classic algorithm is introduced and analyzed. To improve the accuracy of detecing, the research designs a method combined pitch detecting with duration detecting. The simulation shows that it can improve the accuracy of musical fragment recognition.The paper explores representation of music information, and realizes a algorithm of matching, which is tested by Matlab and C program. Based on the algorithm, the whole system in PC is designed and realized. To improve the speed of feature extraction of music, the DSP platform is considered. The thesis discusses the operation of floating-point data on fixed-point DSP, and realizes the key module of feature extraction based on TMS320C6416T.