The Fusion Matching Method for Polyphonic Music Feature Database
This article proposes the fusion matching method for polyphonic music feature database which are extracted from music signal. The best way looking for the song is the tag based retrieval method using metadata like title, singer, lyrics, etc. This is very convenient and powerful way if you have already known about information of contents what you are looking for. But if you do not have any information of the contents, contents based query method might be a plan-B. Query by Singing/Humming (QbSH) is the powerful tool and the best supplemental method looking for song or music over the internet or among huge database. This topic has been researched for a so long time with various solutions. But, there have not been any outstanding solution so far. So we propose the fusion matching method with three matchers against polyphonic music signal in order to improve matching performance. Proposed method is based on Dynamic Time Warp (DTW), Linear Scaling (LS) and Quantized Binary Code (QBcode) and then combines them with fusion score based PRODUCT rule.
KeywordsMIR QbSH DTW LS
- 2.Ghias A et al (1995) Query by humming-musical information retrieval in an audio database. In: Proceedings of ACM Multimedia, pp 231–236Google Scholar
- 3.Roger Jang’s corpus, http://neural.cs.nthu.edu.tw/jang2/dataSet/childSong4public/QBSH-corpus/
- 4.Jang JSR, Lee HR (2008) A General framework of progressive filtering and its application to query by singing/humming. IEEE Trans Speech, Audio Language 2(16):250–258Google Scholar
- 5.Downie JS (2008) The music information retrieval evaluation exchange (2005–2007): a window into music information retrieval research. Acoust Sci Technol 29(4):247–255Google Scholar