Research And Implementation On Content-Based Music Retrieval Technology

With the rapid development of multimedia and computer network technology, the amount of music information rises sharply. How to retrieval the required music accurately and quickly from such vast music data has become a hot spot in the field of content-based music retrieval. Query by Humming becomes one of the main music retrieval methods because of its advantage in conveniences and accurate expression of retrieval requirements. When the user hums a piece of melody, Query by Humming system can search the song from large-scale music database.Focus on key technologies involved in Query by Humming, including: the expression and extraction of music melody feature, the construction of music database, and the approximate matching of music melody, etc. A content-based Querying by Humming system is designed and implemented. Main tasks of the paper are as following:1. Researched the relevant theoretical knowledge of Query by Humming, analyzed the involved key technologies and gave a basic framework of Query by Humming.2. The expression and extraction of music melody feature. Under the study of the basis feature of music and the expression method of music melody, an improved music melody feature extraction algorithm is proposed, which integrated pitch with rhythm feature. It can extract music melody feature more accurately from user's humming input, and have better retrieval accuracy.3. Music retrieval matching algorithm. On the basis of comparing and analysizing the existing algorithm, an improved hierarchical music retrieval matching algorithm is proposed. The experimental results show that the use of this algorithm can get higher success rate, and it can also improve the efficiency of Query by Humming system.4. Based on .Net development platform, a content-based Querying by Humming system is designed and implemented, and then achieved the ideal result.