An algorithm for multi tempo music lyric transcription

Abstract

This paper documents an attempt to create an algorithm for multi-tempo music lyric transcription. This paper reviews music information retrieval as a field of study and identifies music lyric transcription as a subset of the music information retrieval field. The difficulties of music lyric transcription are highlighted and a gap in knowledge in the field is identified. There are no algorithms for music transcription that are applicable to all forms of music; they are usually specialised by instrument or by genre. The author attempts to fill this gap by creating a method for multi-tempo music lyric transcription. The methodology used to achieve this goal is a three-step process of taking audio as input, processing it using the REPET separation technique, and transcribing the separated audio file. The result of this paper was a relative success, with the music being separated successfully and the lyrics being transcribed but with accuracy lost.