Table 1 Word error rates (%) on the Aurora2 task (Results are averaged between 0 and 20 dB SNRs).
From: Histogram Equalization to Model Adaptation for Robust Speech Recognition
Test Sets | Baseline | CMN | CMVN | Multicondition | MLLR | HEQ-FC | HEQ-MA (mean-only) | HEQ-MA (mean & var.) |
|---|---|---|---|---|---|---|---|---|
A |
|
|
|
|
|
| 21.13 | 15.19 |
B |
|
|
|
|
|
| 17.63 | 14.00 |
C |
|
|
|
|
|
| 21.95 | 15.93 |
Average |
|
|
|
|
|
| 19.90 | 14.86 |
