 |
International Workshop on Hands-Free Speech Communication (HSC2001)April 9-11, 2001 |
 |
|
International Workshop on Hands-Free Speech Communication (HSC2001)April 9-11, 2001 |
|
The performance of speech recognition algorithms often degrades drastically when the testing environment is somewhat different from the training conditions. Although many approaches have been proposed to cope with this mismatch problem, learning from extra data collected in operational conditions is always an attractive option if more processing is permitted. Supervised adaptation is the process of learning from an additional set of labeled adaptation data. On the other hand, unsupervised compensation is often referred to as learning from testing data directly. Adaptation and compensation share many similar principles and techniques. We examine some recent advances in this active research area and discuss their applications to improve recognition performance in hands-free communications.
Bibliographic reference. Lee, Chin-Hui (2001): "Adaptation and compensation for speech recognition - learning from extra data to improve robustness", In HSC2001, 27-30.