摘要 |
<p>PROBLEM TO BE SOLVED: To make it possible to provide a user with more comfortable training by previously regulating the speech utterance to the user by a game during training and imparting the utterance for training of a speech recognizing device. SOLUTION: The speech utterance, here words W1 to W3, are given to the user during the training stage and an utterance is given relating to the training procedure of the speech recognizing device to be used in order to make the speech recognition training. In this case, the words W1 and W2 express the words already trained by the speech recognizing device and on the other hand, the speech recognizing device is still required to train the word W3 during this training stage. The words W1 to W3 displayed with respect to the computer game used in this training acts as a mark relating to associated display regions 3 to 5. When a sufficient number of the speech inputs are carried out relating to the word W3 to be trained, a hidden Markov model(HMM) belonging to the word W3 is evaluated in accordance therewith and the set of the HMM model structure and the parameter values belonging thereto is determined.</p> |