90/150 TPSACA를 이용한 패턴분류기의 설계

Abstract

The essential prerequisites of designing the pattern classifier include high throughput, low memory requirements, and low cost hardwired implementation. Several studies have been done on the MACA based pattern classifier. This pattern classifier reduces the problem of complexity in calculating, however, it takes too much time to construct the 0-tree of the MACA. The pattern classifier proposed in this thesis can be a method to solve this problem. We analyze 90/150 TPMACA and design an efficient pattern classifier by synthesizing a number of 90/150 TPMACA. Using this pattern classifier, we can easily find the attractor tree that a pattern could belong. The two stage pattern classifier using 90/150 TPSACA, synthesizing a number of 90/150 TPMACA into a single bit, maps an n-bit patter generated by pattern classifier which 90/150 TPMACA. As a result, the pattern classifier proposed in this thesis will classify a larger test pattern set.