| 摘要 |
PURPOSE:To recognize a large number of vocabulary with a template buffer small in capacity by providing a function to switch a recognition targeted subset at every phase of speech recognition. CONSTITUTION:The device is provided with a subset data base 6 to designate a candidate in the subset in each phase of speech recognition, an accumulation part 7 to store a sound template, a rewritable template buffer 2, and a circuit to transfer a template from the accumulation part 7 to the template buffer 2. Then, the content of the template buffer 2 is rewritten to the template buffer of the candidate of the recognition targeted subset at every phase of speech recognition. In such a way, since it is enough to prepare the capacity of the template butter capable of storing the maximum number of subsets at every phase, it is possible to recognize a large number of vocabulary with the template buffer small in capacity. |
