Design and Implementation of Music Information Retrieval and Gathering Engine (MIRAGE)

Abstract

Query by Singing/Humming (QBSH) is a most natural way for music recognition. A QBSH system can help find songs by matching a part of melody users sing or hum. Many Music Information Retrieval (MIR) techniques have been used to carry out the QBSH goal. In the past, little studies had ever been revealed about how to design and implement a simple QBSH system. In this article, we combine Short-Time-Fourier-Transform, Fourier Transform and Spectrum Analysis techniques at the client end, and Dynamic Time Warping (DTW) and dynamic programming techniques at the server end to implement Music Information Retrieval And Gathering Engine (MIRAGE), which can take input queries issued by users from web-based client-ends toward MIRAGE deployed at the server-end. MIRAGE can identify melody when a user only knows the rhythm of a song without any other information.