Yujiro Harada ¹, Kuniaki Fujimoto ¹, Kei Eguchi ², Masaaki Fukuhara ³, Masahiro Yoshida ³

Affiliations
1 Graduate School of Science and Technology, Tokai University, Kumamoto, JP
2 Department of Information Electronics, Fukuoka Institute of Technology, Fukuoka, JP
3 Department of Embedded Engineering, Tokai University, Tokyo, JP

Abstract

Recently, pattern recognition technologies such as character recognition and finger print recognition are used in various fields. Accordingly, the retrieval technology is becoming important to retrieve the most similar data from a huge database. However, the real-time search is difficult for a software approach, because a digital computer must compare the called data sequentially. To solve this problem, an associative memory has been studied in order to retrieve the reference data which is the most similar to an input data. The associative memory is one of functional memories which is capable of high-speed retrieving for the most similar data by operating parallel. Human brain can retrieve the most similar data to the input data instantly, because it processes information parallel. Accordingly we have focused a neuron CMOS inverter which has characteristics of the brain neuron. In the associative memory, the Hamming distance search circuit is one of the most important circuit to achieve real-time search. In this paper, we propose a wide range Hamming distance search circuit by utilizing neuron CMOS inverters. Unlike conventional circuits, the proposed circuit can retrieve all reference data within a specified Hamming distance.

Keywords

Associative Memory, Hamming Distance, Neuron CMOS Inverter, Time domain.

Full Text