Font Size: a A A

Research On The Technologies Of Content-based Music Retrieval By Humming

Posted on:2008-10-28Degree:MasterType:Thesis
Country:ChinaCandidate:Y R GouFull Text:PDF
GTID:2178360272469302Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
People may have such experiences, when you want to query some music, you have only remember a part of the melody of the music and forget the name, composer, and other information. At this time, it is difficult to use the traditional method of text-based music retrieval to query music. If we take this melody as a input by humming to query the music from the music database, it is regarded as a simple efficient way of music retrieval. The technology of music retrieval by humming is a technology of content-based music retrieval essentially.Melody features extraction and approximate melody match is the core of the technologies on music retrieval by humming .the melody features includes the pitch and the duration of notes. Because of various types of music file format and the different precise and difficulty of extracting the melody features from music of different types of file format, MIDI(Musical Instrument Digital Interface)files which can record the information of the notes accurately are chose as the data of music database. Extract the main melody notes from the MIDI file by calculating and comparing the four features of various MIDI channels, such as the length of channel, the number of overlapping notes, the average pitch and entropy. time domain signal processing technology is used to segment the notes and get the pitch of notes for the humming music, the adaptive dual-threshold segmentation method which based on short-term energy is used to cut notes, the center wave autocorrelation function is used to get the fundamental frequency and the largest weighting method is used to extract the pitch of notes from the fundamental frequency. Finally, the approximate melody Matching method which based on dynamic programming is used to query the music.The different algorithms for the main channel extraction and the different approximate matching algorithms to search the music are compared. The prototype system requires users to hum with "DA DA DA", it can find the music from the music library including 360 MIDI music, if the melody of the hummed music is right.
Keywords/Search Tags:content-based music retrieval, Query By Humming, melody extracting, approximate matching
PDF Full Text Request
Related items