Dynamic Time Warping
Revision as of 21:41, 14 February 2009 by Haspelmath (talk | contribs)
| STUB |
Speech recognition systems are often faced with differences in duration between the stored word templates and the newly arrived realizations of those words. In order to deal with these differences, the recognition system can skip frames in input speech in case these differ too much from the corresponding frames in the template, or it can link certain input frames more than once to the same template frame. This approach is known as Dynamic Time Warping.
