| 摘要 |
PURPOSE:To facilitate the search of the cause of abnormality by converting a voice input signal to a digital sound signal, and storing it in a memory so that even when abnormality, etc., are generated in the processing of a voice analyzing and recognizing device, etc., it can be reproduced. CONSTITUTION:At the time of verifying a recognition algorithm, in the case a fault such as segment abnormality, etc., of an initial end and a terminal of voice data is generated, a control circuit 8 sets a memory 7 to a reproducing state. When the memory 7 is set to the reproducing state, the control circuit 8 instructs a head address for reproducing a digital sound signal to the memory 7. The memory 7 reproduces successively the digital sound signal, while bringing an address to increment by synchronizing with a sampling period of a sample hold 3 from the instructed head address, and supplies its signal to a voice analyzing and recognizing circuit 5. In such a way, when abnormality is generated, the digital sound signal outputted from the memory 7 is supplied to the recognizing circuit 5 and an analyzing operation and a recognizing operation such as a segmentation of a voice and a calculation of a feature parameter, etc., are executed, and the cause of generation of abnormality can be searched. |
