Fuzzy inference chip FZP-0401A based on interpolation algorithm

Naoyoshi Yubazaki*, Masayuki Otani, Akira Muto, Takatsugu Ashida, Jianqiang Yi, Kaoru Hirota, Yan Shi

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)

Abstract

A fuzzy inference hardware which adopts a new linear interpolation algorithm is proposed. Linearity to an antecedent item is proved about Product-Sum composition with limitations to membership functions in the antecedent part. A fuzzy inference calculation algorithm based on Product-Sum-Gravity method is derived by new linear interpolation. In this algorithm, first moment and area data corresponding to the representative points from all situations are stored to memory in advance. Next, fuzzy inference result is interpolated by fired representative point data and distance between observation data and the representative point about each antecedent item. Also, fuzzy inference chip FZP-0401A based on the proposed interpolation algorithm is introduced. The chip supports a fuzzy system with 4 inputs and 1 output of 16 bit. Each input can have 7 labels of membership functions and the output can have 255 labels of membership functions. This chip can process maximally 2401 fuzzy production rules. The size of FZP-0401A is about 120000 gates in gate-array design. The inference speed of the chip is about 0.48 MFLIPS, which does not vary with number of fuzzy production rules, number of labels of membership functions, etc.

Original languageEnglish
Pages (from-to)299-310
Number of pages12
JournalFuzzy Sets and Systems
Volume98
Issue number3
DOIs
Publication statusPublished - 1998
Externally publishedYes

Keywords

  • Fuzzy inference chip
  • Interpolation
  • Look-up-table
  • Product-sum-gravity method

Fingerprint

Dive into the research topics of 'Fuzzy inference chip FZP-0401A based on interpolation algorithm'. Together they form a unique fingerprint.

Cite this