Software Implementation Issues of Existing and New Defuzzification Methods

This paper discusses software implementation issues of different defuzzification procedures in fuzzy systems. Three new defuzzification methods are introduced which are suitable for efficient software and also hardware implementations. A set of seven important existing defuzzification methods are reviewed and compared with these new methods for different software implementation approaches. The C models of all methods are prepared to perform a comprehensive analysis on the output accuracy of different methods. The results prove the superiority of our new proposed methods. In another study, three categories of low-level assembly models are developed for each method to evaluate its software execution time and instruction count when executed on each of three chosen popular processors. Namely, Texas Instruments C6xcopy DSP, Intel's Pentiumcopy IV, and IBM's PowerPC PPC405copy processors are used as the running engines for this comparison. Some accuracy-speed analysis diagrams are then introduced to guide the designers for choosing the defuzzification method which best suites their application requirements.