| Download | - View final version: Joint speaker features learning for audio-visual multichannel speech separation and recognition (PDF, 15.0 MiB)
|
|---|
| DOI | Resolve DOI: https://doi.org/10.21437/Interspeech.2024-1063 |
|---|
| Author | Search for: Li, Guinan; Search for: Deng, Jiajun; Search for: Chen, Youjun; Search for: Geng, Mengzhe1; Search for: Hu, Shujie; Search for: Li, Zhe; Search for: Jin, Zengrui; Search for: Wang, Tianzi; Search for: Xie, Xurong; Search for: Meng, Helen; Search for: Liu, Xunying |
|---|
| Affiliation | - National Research Council of Canada. Digital Technologies
|
|---|
| Format | Text, Article |
|---|
| Conference | Interspeech 2024, September 1-5, 2024, Kos, Greece |
|---|
| Subject | speaker features; zero-shot adaptation; speech separation; speech recognition |
|---|
| Abstract | This paper proposes joint speaker feature learning methods for zero-shot adaptation of audio-visual multichannel speech separation and recognition systems. xVector and ECAPA-TDNN speaker encoders are connected using purpose-built fusion blocks and tightly integrated with the complete system training. Experiments conducted on LRS3-TED data simulated multichannel overlapped speech suggest that joint speaker feature learning consistently improves speech separation and recognition performance over the baselines without joint speaker feature estimation. Further analyses reveal performance improvements are strongly correlated with increased inter-speaker discrimination measured using cosine similarity. The bestperforming joint speaker feature learning adapted system outperformed the baseline fine-tuned WavLM model by statistically significant WER reductions of 21.6% and 25.3% absolute (67.5% and 83.5% relative) on Dev and Test sets after incorporating WavLM features and video modality. |
|---|
| Publication date | 2024-09-01 |
|---|
| Publisher | ISCA |
|---|
| In | |
|---|
| Language | English |
|---|
| Peer reviewed | Yes |
|---|
| Export citation | Export as RIS |
|---|
| Report a correction | Report a correction (opens in a new tab) |
|---|
| Record identifier | bd04a596-9d27-4c63-a31a-7af7ad46aac1 |
|---|
| Record created | 2025-04-10 |
|---|
| Record modified | 2025-04-16 |
|---|
