Digital fuzzy inference system

An if-part membership function calculation circuit for use in a digital fuzzy inference system in which a fuzzy rule is represented by a membership function and a position of a center of gravity of the membership value for each of fuzzy rules is obtained, said calculation circuit comprising:

• first subtracting means for calculating .vertline.xo-xi.vertline. from a point xo of inflection and input data xi;
• multiplying means for calculating k.times..vertline.xo-xi.vertline. from the output of said first subtracting means and a slope k;
• and
• second subtracting means for calculating C-k.times..vertline.xo-xi.vertline. from the output of said multiplying means and a predetermined value C.

An if-part membership function calculation circuit for use in a digital fuzzy inference system in which a fuzzy rule is represented by a membership function and a position of a center of gravity of the membership value for each of fuzzy rules is obtained, said calculation circuit comprising:

• multiplying means for calculating k.multidot.xi from a slope k and input data xi;
• means for setting a product k.multidot.xo of a point xo of inflection and the slope k as a parameter;
• first subtracting means for calculating .vertline.k.multidot.xo-k.multidot.xi.vertline. from the output of said multiplying means and the parameter k.multidot.xo;
• and
• second subtracting means for calculating C-.vertline.k.multidot.xo-k.multidot.xi.vertline. from the output of said first subtracting means and a predetermined value C.

An if-part membership function calculation circuit for use in a digital fuzzy inference system in which a fuzzy rule is represented by a membership function and a position of a center of gravity of the membership value for each of fuzzy rules is obtained, said calculation circuit comprising:

• means for setting a product k.multidot.xo of a point xo of inflection and a slope k as a parameter;
• subtracting means for calculating C-k.multidot.xo from a predetermined value C and the parameter k.multidot.xo;
• multiplying means for calculating k.multidot.xi from input data xi and the slope k;
• and
• add-subtract means for calculating k.multidot.xi-.vertline.C-k.multidot.xo.vertline. when the output of said subtracting means is positive and k.multidot.xi+.vertline.C-k.multidot.xo.vertline. when the output of said subtracting means is negative.

A digital fuzzy inference system for performing fuzzy inference on the basis of input binary data comprising:

• means for outputting a membership value representing an area corresponding to an inferential result for each of fuzzy rules to a plurality of predetermined addresses determined by a fuzzy rule;
• and
• means for calculating a position of a center of gravity of the membership value outputted outputted by said means for outputting;
• wherein said outputting means includes a plurality of calculating means for multiplying the membership value by respective specified parameters in order to divide areas corresponding to the inferential result into plural sub areas which are to be outputted to respective addresses.

A digital fuzzy inference system having a plurality of fuzzy rules comprising:

• fuzzy inference means for performing fuzzy inference on each of the fuzzy rules to provide inferential results;
• first means for calculating a position of the center of gravity of the inferential results;
• second means for digitally recalculating the position of the center of gravity of the inferential results;
• and
• address split designating means for outputting the calculation result of said center-of-gravity position calculating means of a desired resolution;
• wherein said second means provides a digital output, and said address split designating means comprises means for selecting a predetermined number of significant bits of a digital output of said second means.

A digital fuzzy inference system having a plurality of fuzzy rules comprising:

• a plurality of processors each of which being provided for each of fuzzy rules and having an if-part membership function represented by a line having a combination of gradients, two gradients having different gradients connecting to each other at a point-of change, each of said processors comprising:
• means for inputting a plurality of binary data;
• a plurality of multiplier means for obtaining a product of an absolute value of a difference between a point of change and the binary data and a value of the tangent for each of the plurality of binary data, the multiplier means outputting a product and an overflow output;
• a plurality of logical sum gate means connected to outputs of said multiplier means each for receiving the product and the overflow output;
• means for obtaining a maximum value of outputs of said logical sum gate means for each of the binary data;
• means for subtracting the output of said maximum-value calculation means from a maximum value of membership values;
• and
• means for obtaining a then-part output from the output of said subtracting means.

A digital circuit for determining a value of a membership function with respect to an input value, the membership function being represented by C-k.multidot..vertline.xo-xi.vertline. (where C, k, and xo are constants, and xi is the input value), comprising:

• means for adding xo, a complement of the input value xi, and a carry input and outputting a result of addition and a carry output, the carry output being 0 if an overflow does not occur during an addition procedure and 1 if the overflow occurs;
• means for inverting the carry output of said adding means;
• and
• an EX-OR gate for outputting an EX-ORed value of the result of addition and an output of said inverting means, the EX-ORed value being .vertline.xo-xi.vertline..

A digital circuit for determining a membership value with respect to an input value, comprising:

• means for calculating the membership value c-k.multidot..vertline.xo-xi.vertline. (where C, K and xo are constants, and xi is the input value);
• means for determining whether or not a result of a calculation is negative and, if the result is negative, for outputting an underflow signal having a level "O";
• and
• an AND gate coupled to said determining means with respective inputs of said AND gate receiving said result of a calculation and said underflow signal, whereby said AND gate outputs an ANDed-value of result of a calculation result and said underflow signal.

A digital circuit for determining a membership value with respect to an input value, comprising:

• means for calculating the membership value c-k.multidot..vertline.xo-xi.vertline. (where C, K and xo are constants, and xi is the input value);
• an OR gate coupled to receive predetermined upper bits of said membership value for outputting an OR-ed value of said bits;
• and
• a plurality of OR gates coupled to the output of said OR gate and to respectively receive lower bits of said membership value for outputting OR-ed values of said lower bits and the output of said OR gate, wherein said lower bits are bits other than said predetermined upper bits of the membership value.