Intelligent query by humming system based on score level fusion of multiple classifiers

Gi Pyo Nam,Sung-Joo Park,Kang Ryoung Park,Thi Thu Trang Luong,Hyun Ha Nam

doi:10.1186/1687-6180-2011-21

Gi Pyo Nam, Sung-Joo Park + Show 3 more

Open Access

https://doi.org/10.1186/1687-6180-2011-21

Copy DOI

Abstract

Abstract Recently, the necessity for content-based music retrieval that can return results even if a user does not know information such as the title or singer has increased. Query-by-humming (QBH) systems have been introduced to address this need, as they allow the user to simply hum snatches of the tune to find the right song. Even though there have been many studies on QBH, few have combined multiple classifiers based on various fusion methods. Here we propose a new QBH system based on the score level fusion of multiple classifiers. This research is novel in the following three respects: three local classifiers [quantized binary (QB) code-based linear scaling (LS), pitch-based dynamic time warping (DTW), and LS] are employed; local maximum and minimum point-based LS and pitch distribution feature-based LS are used as global classifiers; and the combination of local and global classifiers based on the score level fusion by the PRODUCT rule is used to achieve enhanced matching accuracy. Experimental results with the 2006 MIREX QBSH and 2009 MIR-QBSH corpus databases show that the performance of the proposed method is better than that of single classifier and other fusion methods.

Highlights

With the rapid increase in music data on the Internet, MP3 players, portable media players (PMP), and smart phones, it is more difficult for the users to find the correct file they want
In this paper, we propose a new QBH system that uses the spectro-temporal autocorrelation (STA) method as a pitch extractor and score level fusion of five matchers based on the PRODUCT rule
We normalized the features of the music and humming data through mean-shifting, median and average filtering, and the min-max scaling method to eliminate the surrounding and peak noises that occur during recording

Summary

Introduction

With the rapid increase in music data on the Internet, MP3 players, portable media players (PMP), and smart phones, it is more difficult for the users to find the correct file they want. Waveform is used to adjust the matching results of the global shape [2,8,12] Based on this taxonomy (i.e., note-based and frame-based methods, bottom-up and top-down methods), previous studies can be classified as follows [13,14]. Through the combination of these five classifiers by the score level fusion of the PRODUCT rule based on comparisons of various fusion rules, the performance of the QBH system is greatly enhanced We proved this by comparing the results for the proposed method to those for a single classifier and various other fusion methods.

Proposed method

Method

Local maximum and minimum point-based LS algorithm

Findings

Conclusion