eduzhai > Applied Sciences > Engineering >

JukeBox A Multilingual Singer Recognition Dataset

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: A text-independent speaker recognition system relies on successfully encodingspeech factors such as vocal pitch, intensity, and timbre to achieve goodperformance. A majority of such systems are trained and evaluated using spokenvoice or everyday conversational voice data. Spoken voice, however, exhibits alimited range of possible speaker dynamics, thus constraining the utility ofthe derived speaker recognition models. Singing voice, on the other hand,covers a broader range of vocal and ambient factors and can, therefore, be usedto evaluate the robustness of a speaker recognition system. However, a majorityof existing speaker recognition datasets only focus on the spoken voice. Incomparison, there is a significant shortage of labeled singing voice datasuitable for speaker recognition research. To address this issue, we assemble textit{JukeBox} - a speaker recognition dataset with multilingual singingvoice audio annotated with singer identity, gender, and language labels. We usethe current state-of-the-art methods to demonstrate the difficulty ofperforming speaker recognition on singing voice using models trained on spokenvoice alone. We also evaluate the effect of gender and language on speakerrecognition performance, both in spoken and singing voice data. The complete textit{JukeBox} dataset can be accessed atthis http URL.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...