| 摘要 |
PROBLEM TO BE SOLVED: To smoothly perform 'cap-verses' by preventing the reduction of a voice recognizing rate in the case of performing the 'cap-verses' while using voice recognizing techniques. SOLUTION: When the 'cap-verses' is started, a recognition control part 11c sets a dictionary for cap-verses to a dictionary managing part 11d as the dictionary of a retrieving object. When input voices from a user are converted to time series audio data by an input signal processing part 11a and the same pattern is repeated more than three times in the first section of time series audio data, a recognizing processing part 11b corrects the audio data so that this pattern can appear just once in the beginning of audio data. Further more, the recognized result based on similarity between the pattern of these audio data and the pattern of a recognizing object word stored in the dictionary for cap-verses set by the dictionary managing part 11d is outputted to the recognition control part 11c. The word having final one sound of a preceding word as an initial letter is outputted to a control part 20 as the recognized result by the recognition control part 11c. |
