Font Size: a A A

Music Structure Analysis And Application

Posted on:2007-04-14Degree:MasterType:Thesis
Country:ChinaCandidate:T L ChenFull Text:PDF
GTID:2178360185985757Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Music is one of the key components of Audio and has a lot of semantic information. Music structure is one of the main representation methods of Music piece and one of the key paths of to music comprehension.Most of the Music analysis and Music Information Retrieval (MIR) researches are based on structural information such as label, MIDI , or music notes etc. In this thesis we first improve and design a series of algorithms that based on a completely new view of music structure to automate analysis the Music Structure and label it after a consider comparison and analysis of the related researches before. Then we consider the applications of Music Structure to audio-based MIR and Music Summarization based on the labeled Music Structure information. First we extract the Pitch Class Profile (PCP) feature vector through the analysis of music representation. This feature perfectly combine the frequency in acoustics level and the Temperament in music semantic level, we use the cosine distance of this feature to represent the similarity of two music clips, then we design a group of algorithms that is inspired from the thought of Edit Distance and Dynamic Programming. They segment the feature vectors into groups at first, then through group similarity match, group recurrent detect, merge recurrent group and structure label joined algorithms to complete the music structure label task. Because this is a really new field of research and no good method of evaluation had been finding, we propose a new evaluation method and the results of the experiments show that it is a good method. Then, inspired by the idea of evaluation method we propose a new method and construct a simple Music Structure based Music Information Retrieval system, the experiment results show that it is a perspective method, the system can give a list of retrieval results ordered by structure similarity that extracted from the automated music structure label system build up before, and two same music piece can be identified almost 100%, so it is very useful in MIR and intelligent play list generation. At last we build up a music summarization system based on the structure label system that can spot out the main theme segment, the accuracy is 76.67%, a encouraging result. It is a key application of the music structure analysis.
Keywords/Search Tags:Music Information Retrieval(MIR), Music Structure Analysis, Music Summarization, Edit Distance, Cosine Distance
PDF Full Text Request
Related items