Table 2 The most dominant feature over depth in the DTAM triphone system without gender information
From: Decision tree-based acoustic models for speech recognition
| Depth range | Most dominant feature |
|---|---|
| 1-5 | MFCC static features |
| 6-7 | MFCC first derivatives |
| 8-13 | Right context |
| ≥ 14 | Left context |