Continuous speech separation (CSS) is an approach to handling overlapped speech in conversational audio signals. A real recorded dataset, called LibriCSS, is derived from LibriSpeech by concatenating the corpus utterances to simulate a conversation and capturing the audio replays with far-field microphones.
Variants: LibriCSS
This dataset is used in 2 benchmarks:
| Task | Model | Paper | Date |
|---|---|---|---|
| Speech Recognition | TS-SEP | TS-SEP: Joint Diarization and Separation … | 2023-03-07 |
| Speech Recognition | GSS + Transducer | GPU-accelerated Guided Source Separation for … | 2022-12-10 |
| Speech Separation | Conformer (large) | Continuous Speech Separation with Conformer | 2020-08-13 |
| Speech Separation | Conformer (base) | Continuous Speech Separation with Conformer | 2020-08-13 |
Recent papers with results on this dataset: