In this paper we present a method for smoothing the pitch sequence to remove the outliers caused by pitch tracking method and errors in sung/hummed voice. This approach is used for constructing a query by singing/humming system. After the pitch sequence is smoothed, the continuous pitch contour is calculated. Then, this feature can be used with Dynamic Time Warping (DTW) matching to compute the difference score between the each query and song. Experimental result of this method on TCS Corpus for query by Singing/Humming will be reported and discussed in detail in this paper. © 2008 Springer Berlin Heidelberg.
CITATION STYLE
Nguyen Truong Duc, T., Le Nhat, M., Nguyen Duc Hoang, H., & Vu Hai, Q. (2008). Continuous pitch contour as an improvement feature for music information retrieval by humming/singing. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5351 LNAI, pp. 1086–1091). https://doi.org/10.1007/978-3-540-89197-0_111
Mendeley helps you to discover research relevant for your work.