Full catalogue

SVQTD (Singing Voice Quality and Technique Database) is a classical tenor singing dataset collected from YouTube, it is mainly used to support supervised machine learning performing paralinguistic singing attribute recognition tasks. In SVQTD, there are nearly 4000 vocal solo segments with $4 - 20$ seconds long, totaling 10.7 hours. These segmenets are partitioned from 400 audios of 6 famous tenor arias. Furthermore, each segment is seperrately labeled on seven verbal scales corresponding to seven paralinguistic singing attributes widely used in vocal pedagogy.