| H89-1010 | were performed with and without | triphone cooccurrence smoothing | on all three test sets . These |
| H90-1060 | bins of the discrete densities , | triphone cooccurrence smoothing | \ -LSB- 7 \ -RSB- , has proven |
| H90-1060 | speaker-specific modeling techniques such as | triphone cooccurrence smoothing | and probabilistic spectral transfonnation |
| H89-1010 | adaptation cooccurrence smoothing , and | triphone cooccurrence smoothing | . These methods were presented |
