How to obtain output fuzzy variable with membership values after inference

I'm working on a self organizing fuzzy controller project. The goal is to build a light weight self organizing controller which was first theorized by Procyk in 1977. To implement self organizing algorithm based on fuzzy performance index approach, I need to add the output fuzzy variable of the controller with the performance index fuzzy variable according to fuzzy extension principle. I have built a fuzzy controller using matlab fuzzy toolkit. Now I need to implement a self organizing fuzzy system which can modify the rule base of the latter. Is there a way to obtain the fuzzy output from the inference engine before de-fuzzification? If fuzzy output is available, using fuzzy extension principle the correction could be applied to the output variable, thereby modifying the entire rule base and input variable definition.

I'm using simple isosceles triangle memberships for every fuzzy variable in normalized fuzzy universe. The fuzzy system is of Mamdani type.