Font Size: a A A

Retrieval Based On Humming The Music

Posted on:2009-02-06Degree:MasterType:Thesis
Country:ChinaCandidate:X H SongFull Text:PDF
GTID:2208360245479070Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With rapid development of multimedia technology, a large number of online music can be obtained from the Internet. Now, people are no longer satisfied with searching music based on text information such as music name and singer. Especially, with various kinds of music data, people might only remember the tune of a song, or perhaps a fragment of it. How to quickly and effectively retrieve music data by melody information has become a prominent problem. In this dissertation, we study and explore the music retrieval based on hummed melody.There exists some kind of randomness when people hum a song. Moreover, women's pitch is usually higher than men's pitch. The pitches of different people are distinct, even for the same person the tone at different time is distinct. Humming the song rightly is difficult. People may hum the tune higher or lower, or sometimes insert or delete some notes. And out of tone is sporadic.After a lot of experiments, we found that the pitch interval is unchanged and the duration rate is same for the same music, although the pitch changed various during the humming.Based on above analysis, this dissertation used the pitch interval and the duration as the music feature vector to avoid the pitch changing. The Continuous Hidden Markov Model, whose output sequence is not limited several values but continuous values, is used for music modeling to find the probability of any pitch interval so as to solve the out of tone problem . At the same time, hidden states, describing the inserting or deleting notes situation, were introduced into the CHMM in order to reduce the query error. For improving the pitch and duration express accuracy and enhance the matching accuracy, we compute the fundamental period by AMDF with a dynamic threshold, and then we did some kind of smoothing to the fundamental period. Duration approximate, which contains rhythm information, was used together with CHMM when find matching music.
Keywords/Search Tags:melody, music retrieval, note segmentation, feature extraction, HMM
PDF Full Text Request
Related items