摘要 |
PROBLEM TO BE SOLVED: To improve recognizable precision in an input device not premising an uttered voice of sound and to shorten a recognition time by executing the clustering processing of major grouping processing of an uttered voice syllable based on a measuring value of a single time while calling. SOLUTION: The measuring data of the displacement data or/and the speed data of every measuring point are gained (S101). The time that a vector length becomes maximum and the time tv after 0.12 sec. elapses are obtained from the measuring data, and the vector in the time tv is obtained (S102, 103). The clustering, that is, the bowel, the contracted sound recognition processing is performed, and the major grouping of the uttered voice syllable is performed based on the shortest distance standard pattern, and the processed result of the clustering are obtained (S104, 105). The time from the time tv gaining the clustering data to tv-1 before 1 sec. is specified as a consonant recognition range (S106). The uttered voice recognition of minor grouping among the grouping major grouped by a dynamic programming method is performed, and the recognition result is obtained (S107, 108).
|