| 摘要 |
PURPOSE:To enable speech recognition as to many and unspecified persons by performing pattern matching processing so as to control adaptively the inclination and format information of a frequency structure that a speech signal has. CONSTITUTION:A control circuit 128 reads variables inputted from data buses 104 and 105 and time-series data on constant out to a memory 127 and sends them out to data buses 100-103. The data on the buses 100-103 are stored in registers 106-111 and outputted. Then specific arithmetic operation is carried out based upon a real root on a Z plane corresponding to the inclination of the frequency structure of the speech signal and a conjugate complex root corresponding to a speech format through multipliers 118 and 119, registers 114 and 115, and an arithmetic logic computing element 120. Consequently, the pattern matching is performed by pipeline processing to accurately perform the speech recognition as to many and unspecified persons. |
