Multi-speaker ASR on Libri2Mix mixboth
19.6WERUME ASR init. (0.33, 0.33, 0.34) w/ RWSE
Evaluation Results
| Method | Links | |
|---|---|---|
| UME ASR init. (0.33, 0.33, 0.34) w/ RWSETraining Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max, Initialization=ASR init., Training weights (λasr, λdiar, λsep)=(0.33, 0.33, 0.34), RWSE=true2025.08 | 19.6 | |
| UME (0.33, 0.33, 0.34) + ASR init.Training Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max, Initialization=ASR init., Training weights (λasr, λdiar, λsep)=(0.33, 0.33, 0.34)2025.08 | 21.1 | |
| UME (0.1, 0.1, 0.8)Training Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max, Training weights (λasr, λdiar, λsep)=(0.1, 0.1, 0.8)2025.08 | 22.4 | |
| UME (0.33, 0.33, 0.34)Training Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max, Training weights (λasr, λdiar, λsep)=(0.33, 0.33, 0.34)2025.08 | 22.7 | |
| Multi-speaker AED [9] (reproduced)Training Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max2025.08 | 24.4 | |
| UME (λasr = 1.0)Training Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max, Training weights (λasr, λdiar, λsep)=(1.0, 0, 0)2025.08 | 25 | |
| UME (0.1, 0.1, 0.8) w/ weighted sumTraining Set=LibriMix, Training hours=460 hours, Mix Type=mixboth: speech + noise, Overlap Mode=max, Training weights (λasr, λdiar, λsep)=(0.1, 0.1, 0.8), Weighted Sum=true2025.08 | 25.5 |