|
In this dissertation we present a series of studies on fast multiple candidate search algorithm for Mandarin speech recognition. The scope of the studies in this dissertation covers the tasks of isolated Mandarin syllable recognition, continuous (connected syllable) Mandarin speech recognition and isolated large-vocabulary Mandarin polysyllabic word recognition. The algorithms we proposed not only have a better efficiency in searching for top-1 candidate than the normal search algorithm, but also are able to get top-N candidates with a very little computational overhead. These fast multiple candidate search algorithms can also be applied to the tasks of isolated large vocabulary Mandarin polysyllabic word recognition and keyword spotting to get a better efficient search.
|