Studying the effects of bass estimation for chord segmentation in pop-rock music

TitleStudying the effects of bass estimation for chord segmentation in pop-rock music
Publication TypeMaster Thesis
Year of Publication2014
AuthorsCapablo-Riazuelo, U.
AbstractIn this thesis we present the research work we have carried out on bass line extraction and transcription and audio segmentation for chord estimation. First, we introduce the task and we define important musical concepts to be taken into account. Next, we review the scientific background to our work, both in the chord estimation and the bass extraction/transcription tasks. We then present a set of modifications to the Essentia's predominant melody algorithm to adapt it for bass line estimation. Bass information along with beat positions are then used to propose a novel type of audio segmentation for chromagram smoothing related to the chord estimation problem. Next, we present the evaluation methodology, music collections and metrics used in our research, followed by the evaluation results. The results show a considerable improvement in the bass extraction task by using our approach and promising results in the bass transcription task. They also show very promising results regarding our novel audio segmentation method for chromagram smoothing, compared to the beat-synchronous chromagram approach used by current state-of-the-art algorithms. The thesis concludes with the contributions of our dissertation, the challenges found during the research process and the future work.
intranet