Abstract | ||
---|---|---|
In this paper, we present a method for incremental on-line adaptation based on feature space Maximum Likelihood Linear Regression (FMLLR) for telephony speech recognition applica- tions. We explain how to incorporate a feature space MLLR transform into a stack decoder and perform on-line adapta- tion. The issues discussed are as follows: collecting adapta- tion data on-line and in real time; mapping adaptation data from previous feature space to the present feature space; and smoothing adaptation statistics with initial statistics based on original acoustical model to achieve stability. Testing results on various systems demonstrate that on-line incremental FM- LLR adaptation could be an eective and stable method when the adaptation statistics are mapped and smoothed. |
Year | Venue | Keywords |
---|---|---|
2002 | INTERSPEECH | feature space,real time,speech recognition |
Field | DocType | Citations |
Feature vector,Pattern recognition,Computer science,FMLLR,Speech recognition,Smoothing,Feature (machine learning),Maximum likelihood linear regression,Artificial intelligence,Telephony | Conference | 15 |
PageRank | References | Authors |
0.93 | 6 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Yongxin Li | 1 | 30 | 3.03 |
H. Erdogan | 2 | 589 | 55.11 |
Yuqing Gao | 3 | 480 | 53.03 |
Etienne Marcheret | 4 | 100 | 11.15 |