Abstract | ||
---|---|---|
We present a new predictive compensation scheme which makes no assumption on how the noise sources alter the speech data and which do not rely on clean speech models. Rather, this new scheme makes the (realistic) assumption that speech databases recorded under different background noise conditions are available. The philosophy of this scheme is to process these databases in order to build a "tool" which will allow it to han dle new noise conditions in a robust way. We evaluate the perfor- mances of this new compensation scheme on a connected dig- its recognition task and show that it can perform significant ly better than multi-conditions training, which is the most wi dely used techniques in these kind of scenarios. |
Year | Venue | Keywords |
---|---|---|
2003 | INTERSPEECH | speech recognition |
Field | DocType | Citations |
Background noise,Computer science,Speech recognition,Artificial intelligence,Connected digits,Machine learning | Conference | 0 |
PageRank | References | Authors |
0.34 | 8 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Khalid Daoudi | 1 | 145 | 23.68 |
Murat Deviren | 2 | 26 | 4.65 |