| Summary: | 碩士 === 輔仁大學 === 資訊工程學系 === 97 === In content-based music information retrieval field, because Internet development and the MP3 digital music to grow fast. Music quantity is getting more and more huge, and has many unknown music. Therefore, provides the music information and the management service is an important research subject.
In this paper, we design music identification and management system, system discriminate and compile music information, and music classification.
Out Music data form is MP3 and WAVE. We extract music feature of Mel-scale frequency cepstrums coefficients. We proposed Bisecting K-means clustering algorithm to clustering for music feature and clustering, and chooses music reference feature from several clustering. Music reference feature according to the k-nearest neighbor classifier method is used to measure the similarity between music objects. Finally, we implement a prototype to demonstrate our approach.
|
|---|
