摘要 |
PURPOSE:To recognize a speech which is voiced successively by one >=2 words with small storage capacity. CONSTITUTION:This speech recognizing method is provided with a primary collation part 1 which analyzes the input signal of the speech voices successively by one word or >=2 words, collates all the words with standard patterns, and outputs the result as a sequence of similarity and a secondary collation part 2 which handles the sequence of similarity outputted by the primary collation part as parameters for recognition and collates with previously generated standard patterns for secondary collation; and the similarity sequence outputted as the result of the primary collation is collated by the secondary collation part with all the standard patterns to be recognized which are generated previously and the speech of one word or >=2 successively voiced words is recognized by using the result of the collation. |