| This is the baseline Kaldi recipe for unaccompanied karaoke singing using DSing corpus. | |
| The DSing corpus is derived from the Smule Sing! 300x30x2 karaoke corpus [1]. | |
| It possesses three training sets: DSing1 with 15 hours of GB accent, | |
| DSing3 with 45 hours of GB+US+AU accents, and DSing30 with mix of accents from 30 countries. | |
| It was prepare as a speech recognition corpus by Gerardo Roa Dabike. | |
| The recipe is in s5/ | |
| [1] Smule Sing! 300x30x2 Dataset, “https://ccrma.stanford.edu/damp/” accessed September 2018. |