A Lyrics-matching Qbh System for Inter- Active Environments

Query-by-Humming (QBH) is an increasingly prominent technology that allows users to browse through a song database by singing/humming a part of the song they wish to retrieve. Besides these cases, QBH can also be used to track the performance of a user in applications such as Score Alignment and Real-Time Accompaniment. In this paper we present an online QBH algorithm for audio recordings of singing voice, which uses a multi-similarity measurement approach to pinpoint the location of a query within a musical piece taking into account the pitch contour, phonetic content and RMS energy envelope. Experiments show that our approach can achieve 75.4% Top-1 accuracy in locating an exact melody from the whole song, and 57.8% Top-1 accuracy in locating the phrase that contains the exact lyrics – an improvement of 170% over the basic pitch contour method. Average query duration is 6 seconds while average runtime in MATLAB is 0.8 times the duration of the query.