| 摘要 |
PROBLEM TO BE SOLVED: To improve speech recognizing performance without changing a recognition engine. SOLUTION: A speech data generation part 14 generates a plurality of pieces of speech data whose non-speech sections start at different position from speech data as an object of speech recognition, the recognition engine 17 performs speech recognition by using the respective pieces of speech data, and a totaling/comparison part 19 outputs the most recognition result among a plurality of obtained recognition results. COPYRIGHT: (C)2004,JPO&NCIPI
|
