The query by humming method based on cluster analysis and dynamic time warping
In the modern information society, multimedia information has become a main component of the information superhighway. Users often query the multimedia information according to their need. Traditional multimedia retrieval based on text (such as Baidu, Google) has its inherent limitations. Because people are unable to retrieve multimedia files without key words, content-based retrieval was proposed in response to the proper time and conditions. Of many music retrieval methods, the music retrieval based on melody matching is a popular research direction in recent years. According to the musical characteristics such as melody and rhythm, we can efficiently achieve matching. Query by humming (QBH) is considered as a content-based music retrieval way that is convenient and easily accepted by users. Namely inquirers record a music melody through microphone and then can retrieve similar music on computer. Many research institutions such as MIT and University of Southern California have done some research about content-based audio retrieval, and have made great progress in many fields such as QBH, audio classification, and audio structured . Currently, the Muscle Fish Company has developed a commercial audio retrieval engine based on audio perceptual characteristics. Siri voice recognition system in Apple products has a high utilization rate and a wide range of user . A few Chinese web hum retrieval systems have come into use, but retrieval effect is not satisfactory. Current research results and applications show that QBH remains to be further perfected.