| 摘要 |
PROBLEM TO BE SOLVED: To realize a higher recognition performance by calculating a plurality of kinds of intermediate recognition results and by determining a final recognition result from the recognition results. SOLUTION: A voice is inputted from a microphone 1 and is transmitted to a plurality of recognition processing portions 2(a) to 2(z). In each recognition means included in each recognition processing portions 2, different recognition processings are performed, and intermediate recognition results are outputted. The outputted recognition result include, for example, an alignment of similarities of pattern matching results in the recognition processing. This outputted intermediate recognition result is transmitted to a recognition result determining portion 3, and the recognition result determining portion 3 determines a final recognition result. Thus, a detail recognition result can be calculated using the intermediate recognition result derived in each recognition processing, and therefore, recognition performance can be improved. |
