In order to improve the performance of speech recognizers, particularly in degraded environmetns, it may be beneficial to integrate use of temporal information. As literature has shown that human listeners are able to use temporal cues in speech recognition tasks, this study examiens algorithms for extraction of temporal cues in a speech signal. The task under analysis is the location of landmarks to direct further analysis in a knowledge-based speech recognition sytem. I nthis study, a set of robust landmark types were located with an accuracy of 82.5
展开▼