| Summary: | 碩士 === 輔仁大學 === 資訊工程學系 === 95 === The research about content-based music information retrieval has already proposed a lot of perform algorithms in recent years, and the research issues develops the polyphonic music database at the most from the monophonic music database. The ways to search from polyphonic music database by matching melody are to calculate the absolute position of pitch and duration in extant many polyphonic music information retrieval. This kind of method can look out for the similar answer, but it has neglected music structural that is the concept of voice. The concept of what is called voice is when music melody of many sound is going on, the song consists of a lot of monotony music melodies or have one theme. So if inquire the result published is distributed in different voice, its similar degree will be relatively low. This paper deals with the monophonic music matching to the polyphonic music database. First of all, analyze to the distribution of the voice structure , pass the dissimilarity and voice value, will inquire that does more rational sequencing in result. Finally, we combine and function as two systems in fact according to algorithms of performing proposed: Make Voice System(MVS) and Query and Ranking System(QRS).
|
|---|
