WSEAS Transactions on Systems
Print ISSN: 1109-2777, E-ISSN: 2224-2678
Volume 15, 2016
Query By Singing/Humming System Using Segment-Based Melody Matching for Music Retrieval
Authors: ,
Abstract: Query By Singing/Humming is a melody recognition system for music retrieval by using a singing or humming query. A segment-based melody matching approach is proposed to solve the problems of puff noise and inconsistent tempo, and to lower the computational complexity of traditional Linear Scaling method in Query By Singing/Humming system. The query and midi in database are separated into several segments by four methods we proposed, namely, “Cross n Semitones”, “Moving Average”, “Composite Moving Average”, and “Combination of Cross n Semitones and Composite Moving Average”. Linear scaling is then applied in each corresponding segment. The top 10 recognition rate of our method can reach 70.6%. In addition, from our examples, it shows that segment-based method can solve the problems of puff noise and inconsistent tempo better than traditional global Linear Scaling. The segment-based methods also reduce the computation complexity by jumping to the possible segment boundary and using the segment ratio as scaling ratio, instead of comparing frame by frame from the very beginning and trying different scaling ratio in traditional Linear Scaling method.
Search Articles
Keywords: Query By Singing/Humming, Music Retrieval, Melody Matching, Segment-Based, Linear scaling, Moving Average
Pages: 157-167
WSEAS Transactions on Systems, ISSN / E-ISSN: 1109-2777 / 2224-2678, Volume 15, 2016, Art. #17