Singing and narration are fundamental forms of vocal expression. Listener-evaluated data is essential for developing AI systems that better capture the nuances of vocal expression and resonate with listeners perceptually and emotionally. In this paper, we present a dataset with 4300 ratings of 940 recordings of amateur karaoke singing and audiobook narration, annotated by 86 participants. Participants rated multiple aspects of audio quality and vocal performance including skill, likability, passion, sincerity, and intelligibility, among others. We release this dataset alongside baseline analyses comparing ratings of singing and audiobook narration, as well as ratings of clean excerpts versus those with added degradations. We also release a linear regression model that estimates listener ratings from audio. This dataset will serve as a valuable resource for future research in music and speech, with applications in intelligent music production, vocal audio effect recommendation, audiobook editing, content personalization, and automated assessment of singing and speaking voices, to name a few.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=23008
(2225KB)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
