One of the difficulties in Language Recognition is the variability of the speech signal due to speakers and channels. If channel mismatch is too big and when different categories ...
For spoken document retrieval, it is very important to consider Out-of-Vocabulary (OOV) and mis-recognition of spoken words. Therefore, sub-word unit based recognition and retriev...
Face-to-face meetings usually encompass several modalities including speech, gesture, handwriting, and person identification. Recognition and integration of each of these modaliti...
Michael Bett, Ralph Gross, Hua Yu, Xiaojin Zhu, Yu...
Accurate unsupervised learning of phonemes of a language directly from speech is demonstrated via an algorithm for joint unsupervised learning of the topology and parameters of a ...
—Traditionally, HMM-based approaches to online Kanji handwriting recognition have relied on a hand-made dictionary, mapping characters to primitives such as strokes or substrokes...