Hardware Implementation of IDS Method

Please enable javascript in your browser.

Tarkhan, Mahdi | 2009

625 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 39812 (05)
University: Sharif University of Technology
Department: Electrical Engineering
Advisor(s): Bagheri Shouraki, Saeed
Abstract:
ALM (Active Learning Method) is based on fuzzy concepts, striving for human-like information processing. Typical fuzzy methods for control and modeling applications rely on linguistic processing with membership functions of the kind in fuzzy rules. ALM, however, processes multiple features of a target by converting each into patterns. This simulates the human approach to the acquisition of knowledge from complex source targets in that it focuses on each feature of the target and processes that information in pattern-like images rather than utilizing numerical interpretations. In ALM, the ink drop spread (IDS) method has been proposed as a means of obtaining useful information as pattern data. An ALM system has plural IDS processing blocks. IDS demands high-speed processing and large memory, and the number of blocks increases extremely according to system complexity. In addition to these requirements, for the purpose of developing an ALM hardware system as a robust soft computing architecture, design of a dedicated hardware for IDS is required. In this thesis a new hardware implementation is proposed for IDS method that is at least 10 times faster than the best hardware solution developed yet.
Keywords:
Active Learning ; Ink Drop Spread (IDS)Operator ; Hardware Implementation ; Analog Memory

No TOC