| 摘要 |
<p>A speech recognition apparatus, for recognizing prede termined speech data sequence, comprises a speech recognition neural network unit (30), to which speed data sequence (100) is to be inputted as feature vectors (110) from a feature extracting unit (10), for performing a speech recognizing action whether or not the input speech data sequence matches with predetermined object speech data sequence, a recognition signal detecting unit (30) for outputting a reset instruction signal (130) each time the speech recognition signal (120) is outputted from the neural network network unit, and an internal state value setting unit (40) for resetting the neural network unit and for setting the internal state value to an internal value each time the reset instruction signal is outputted. Since the neural network unit is reset each time the speech recognition signal is outputted, accurate detection can be achieved even when speech data sequence to be recognized is inputted successively. <IMAGE></p> |
