摘要 |
a) extracting feature vector, if speech is input; b) obtaining probability for the feature vector extracted and a feature vector from each model distribution, and obtaining state sequence corresponding to the probability; c) obtaining probability for maintaining time between states from the state sequence; d) obtaining total probability value of a model for an input, from the probability for the maintaining time and the probability for the feature vector; e) if the total probability value is over a maximum value, replacing the maximum value with the total probability and a model with the model; f) if the total probability value is less than the maximum value, increasing the model until the model becomes a mutual information measure, to thereby return to the step b) for obtaining the probability for the feature vector and the state sequence. |