Table 6 Results obtained with WRE approach compared with the baseline. Command Error Rate depending on the weight reestimation rules (MLE, FFDW and FD) without ULT. 11 136 tests performed on VODIS corpus (noisy).
From: Compact Acoustic Models for Embedded Speech Recognition
WRE | Baseline | |||
|---|---|---|---|---|
MLE | FDW | FD | ||
Very compact model | 6.05% | 8.54% | 5.99% | 5.80% |
Compact model | 5.15% | 7.50% | 5.15% | 4.80% |