Humming note segmentation method in Query-by-Humming

Abstract

Humming note segmentation is an important step in Query-by-Humming (QBH) and the accuracy of note segmentation plays an important role in improving the accuracy of the melody matching in music retrieval system. According to the properties that the voice was not stable, a note segmentation method based on dynamic threshold was put forward in this paper to improve the accuracy of the humming note segmentation. The experiment results showed that the dynamic threshold method can effectively improve the accuracy of retrieval, and it provides an idea to improve the effect of QBH.