Table 1 PER [%] for PLP-based system for progressively refined AM
From: Advanced acoustic modelling techniques in MP3 speech recognition
Data | Baseline | LDA | SAT | SGMM | MMI |
|---|---|---|---|---|---|
Raw | 17.9 | 16.1 | 12.3 | 10.2 | 7.2 |
128k | 18.8 | 17.1 | 13.5 | 10.8 | 7.9 |
32k | 19.3 | 17.3 | 13.2 | 11.0 | 8.2 |
28k | 19.6 | 17.7 | 13.5 | 11.3 | 8.6 |
24k | 20.7 | 18.7 | 14.1 | 11.9 | 9.3 |
20k | 23.9 | 21.1 | 15.5 | 12.9 | 10.5 |
16k | 36.2 | 30.4 | 18.6 | 15.5 | 13.3 |
12k | 62.5 | 52.9 | 26.8 | 22.2 | 20.2 |