| 摘要 |
In a voice recognition system which seeks a plurality of prospective recognition results with respect to at least two input voices and which select each one prospective recognition result among respective a plurality of prospective recognition results with respect to the two input voices, so that an optimized sequence of recognition results is obtained as a whole, primary and secondary language processing are simultaneously carried out by the use of dynamic programming on a primary prospective recognition result of an input voice preceding an appointed input voice and on a secondary prospective recognition result of another input voice following the appointed input voice, respectively, a cumulative score of an optimized path is calculated on each of appointed prospective recognition results with reference to results of the primary and the secondary language processing to sort the appointed prospective recognition results of the appointed voice in optimized order.
|
