| 摘要 |
PURPOSE:To enhance a recognition rate by recognition processing and to reduce substantially the number of standard pattern framing steps by generating automatically an excellent standard pattern from plural learning patterns. CONSTITUTION:An expansion and contraction normalizing means 5 inputs the 1st temporary standard pattern and an average mapping function as signals 22 and 24, respectively, executes nonlinear normalization and outputs the 2nd temporary standard pattern. A mapping function generating means 6 sequentially inputs the 2nd temporary standard patterns and learning patterns as signals 25 and 21, respectively, and outputs the 2nd mapping function. The 2nd expansion and contraction normalizing means 7 sequentially inputs a learning pattern and the 2nd mapping function as the signal 21 and a signal 26, respectively, and outputs an expansion and contraction normalized pattern in correspondence to the learning pattern. An average pattern generating means 8 obtains a weighted average of all said expansion and contraction normalized patterns to be sequentially inputted as a signal 27, outputs it as an average pattern and sets it to a standard pattern.
|
