Automatic Detection of Flamenco Singing Vocal Resources

TitleAutomatic Detection of Flamenco Singing Vocal Resources
Publication TypeMaster Thesis
Year of Publication2019
AuthorsLuque, S. R.
AbstractThe phrasing in flamenco singing displays a unique richness and complexity that can be observed on various levels: timbre-related features, dynamics, rhythm variations, metrics. In this project we focus on a set of common vocal resources of flamenco singing, including them as part of the timbre-related features. This set contains four different types of resources: ornaments, effects, attacks and vibratos, as specified in [1]. In this thesis we aim to build a system for the automatic identification of these vocal resources in flamenco recordings and for its visualization. Given the lack of annotated material in this domain, we present a dataset which includes both monophonic and polyphonic audio files as well as high quality isolated examples performed by professional flamenco singers. We propose an annotation of these vocal resources including identification and temporal location. Using this dataset we aim to identify our targets in monophonic recordings of flamenco professional singers. This will be based on f0 and timbre descriptors. We expect to build a system capable of optimizing the learning and teaching flamenco singing process, by integrating automatic singing analysis tools for visualizing the most important aspects of vocal techniques in singers.
KeywordsFlamenco Singing, music information retrieval, Pattern Mining