Improved Humming Music Retrieval Method Based on Wavelet Transformationand Dynamic Time Warping

摘要

This paper proposes a method, which makes use of theband-pass property of the wavelet transform, combining thewavelet analysis with autocorrelation function to extract melodycharacters from the humming signal. After the wavelettransform, the humming signal is weighed, and then combinedwith auto-correlation function for signal analysis. The weightedfactors in the design, gives full consideration about that thesound of men and women with different frequency range. Thispaper adopts the improved DTW algorithm to match melodies,and presents one kind of music melody representationirrespective with tone. While carrying through the DTWalgorithm, it can avoid the up and down transition of tone andalso reduce the calculation of melody matching.