| 摘要 |
PURPOSE:To reduce the collating range of voice recognition and to conduct a high speed collation by employing a sentence structure analysis section which is simple in its mechanism and has a smaller amount of storage. CONSTITUTION:A filter bank 1 conducts a frequency analysis of an inputted voice. An analysis section 6 analyzes the voice symbol column using a grammar. A collating section 2 decides the class to which the inputted voice belongs. An input section 7 stores an ending symbol and two location numbers. A chart section 5 stores the ending symbol or a non-ending symbol. A grammar section 3 stores a context free grammar. An operation list section 8 stores an LR list. Having the above constitution, a specific class is introduced as the collating object of an unknown inputted voice. |
