Fuzzy inference system and a pattern input type membership value generator

Claim 2

A fuzzy inference system, comprising: a rule memory having a fuzzy rule stored therein, the fuzzy rule including an input variable and a distinction code identifying whether the input variable is a scalar quantity or image pattern data; first adaptation degree operation means for generating a first adaptation degree representing a membership function value corresponding to the input variable when the input variable is a scalar quantity; second adaptation degree operation means for generating a second adaptation degree representing a pattern category value corresponding to the input variable when the input variable is image pattern data; first selector means for selecting one of said first and second adaptation degrees in accordance with said distinction code included in the fuzzy rule; and inference operation means for applying a predetermined operation to the one of the first and second adaptation degrees selected by said first selector means to generate a fuzzy inference output.

Claim 3

A fuzzy inference system according to claim 1, further comprising second selector means for selecting an input signal designated by the input variable included in the fuzzy rule, the second selector means selecting the input signal from among a plurality of input signals provided by a plurality of sensors.

Claim 4

A fuzzy inference system according to claim 1, wherein said input variable is image pattern data from an imaging device.

Claim 5

An antecedent processing device for processing an input variable to generate an antecedent adaptation degree in accordance with a previously set fuzzy rule comprising a distinction code identifying whether the input variable is a scalar quantity or image pattern data, wherein the input variable is expressed using a membership function when the input variable is a scalar quantity and is expressed using a pattern category when the input variable is image pattern data, the device comprising: first adaptation degree operation means for generating a first adaptation degree representing a membership function value corresponding to the input variable to express a scalar quantity about a membership function designated by the fuzzy rule when the input variable is a scalar quantity; second adaptation degree operation means for generating a second adaptation degree representing a pattern category value corresponding to the input variable to express pattern data about a category designated by the fuzzy rule when the input variable is image pattern data; first selector means for selecting one of said first and second adaptation degrees in accordance with said distinction code included in the fuzzy rule; and operation means for applying a predetermined operation to the one of the first and second adaptation degrees selected by said first selector means to generate an adaptation degree of antecedent for the fuzzy rule.

Claim 6

An antecedent processing device for processing an input variable to generate an antecedent adaptation degree in accordance with a previously set fuzzy rule comprising a distinction code identifying whether the input variable is a scalar quantity or image pattern data, wherein the input variable is expressed using a membership function when the input variable is a scalar quantity and is expressed using a pattern category when the input variable is image pattern data, the device comprising: first adaptation degree operation means for generating a first adaptation degree representing a membership function value corresponding to the input variable to express a scalar quantity about a membership function designated by the fuzzy rule when the input variable is a scalar quantity; second adaptation degree operation means for generating a second adaptation degree representing a pattern category value corresponding to the input variable to express pattern data about a category designated by the fuzzy rule when the input variable is image pattern data; first selector means for selecting one of said first and second adaptation degrees in accordance with said distinction code included in the fuzzy rule; and operation means for applying a predetermined operation to the one of the first and second adaptation degrees selected by said first selector means to generate an adaptation degree of antecedent for the fuzzy rule.

Claim 7

An antecedent processing device according to claim 4, wherein said input variable is image pattern data from an imaging device.

Claim 8

An adaptation degree generating device for developing an adaptation degree in accordance with a fuzzy rule expressed using a category of pattern about an input variable of image pattern data, comprising: a standard pattern memory for previously storing standard image pattern data for each of a plurality of categories; distance computing means for computing a distance between input image pattern data and standard pattern data of a category designated by a fuzzy rule of said memory; and adaptation degree generating means for converting the distance generated from said distance computing circuit means to an adaptation degree, said adaptation degree is a value "1" when the distance is zero and decreases as the distance increases.

Claim 9

An adaptation degree generating device for developing an adaptation degree in accordance with a fuzzy rule expressed using a category of pattern about an input variable of image pattern data, comprising: a standard pattern memory for previously storing standard image pattern data for each of a plurality of categories; distance computing means for computing a distance between input image pattern data and standard pattern data of a category designated by a fuzzy rule of said memory; and adaptation degree generating means for converting the distance generated from said distance computing circuit means to an adaptation degree, said adaptation degree is a value "1" when the distance is zero and decreases as the distance increases.

Claim 10

An adaptation degree generating device according to claim 6, wherein said input variable of image pattern data is input from an imaging device.

Claim 11

A pattern input type of membership value generating device, comprising: pattern data input means for inputting image pattern data; standard pattern data generating means for converting a category related to image pattern data entered by said pattern data input means into standard pattern data about said category; distance computing means for computing a distance between input image pattern data entered through said pattern data input means and standard pattern data generated by said standard pattern data generating means; and adaptation degree generating means for converting the distance generated from said distance computing means to an adaptation degree, said adaptation degree is a value "1" when the distance is zero and decreases as the distance increases.

Claim 12

A pattern input type of membership value generating device, comprising: pattern data input means for inputting image pattern data; standard pattern data generating means for converting a category related to image pattern data entered by said pattern data input means into standard pattern data about said category; distance computing means for computing a distance between input image pattern data entered through said pattern data input means and standard pattern data generated by said standard pattern data generating means; and adaptation degree generating means for converting the distance generated from said distance computing means to an adaptation degree, said adaptation degree is a value "1" when the distance is zero and decreases as the distance increases.

Claim 13

A membership value generating device according to claim 8, wherein said standard pattern data generating means generates a feature vector of standard pattern data, and said distance computing circuit means extracts a feature quantity of the input image pattern data, generates a feature vector based on the extracted feature quantity, and computes a distance between the feature vector of the standard pattern data and the feature vector of the input image pattern data.

Claim 14

A membership value generating device according to claim 8, wherein said distance computing means executes a pattern matching of said standard pattern data with said input image pattern data to compute said distance in accordance with a degree of inconsistency between said standard pattern data and said input image pattern data.

Claim 15

A pattern input type of membership value generating device according to claim 8, wherein said pattern data input means further comprises an imaging device.

Claim 16

A fuzzy inference method, comprising the steps of: expressing an input variable of scalar quantity using a membership function and an input variable of image pattern data using a pattern category; previously storing into a rule memory a fuzzy rule comprising a distinction code identifying whether the input variable is a scalar quantity or image pattern data; reading the fuzzy rule from said rule memory; generating a first adaptation degree of the input variable to express scalar quantity about a membership function designated by said read fuzzy rule when the input variable of the rule is a scalar quantity; generating a second adaptation degree of the input variable to express pattern data about a pattern category designated by said read fuzzy rule when the input variable is image pattern data; and applying a predetermined operation to one of said first and second adaptation degrees to generate a fuzzy inference output.

Claim 17

A fuzzy inference method, comprising the steps of: expressing an input variable of scalar quantity using a membership function and an input variable of image pattern data using a pattern category; previously storing into a rule memory a fuzzy rule comprising a distinction code identifying whether the input variable is a scalar quantity or image pattern data; reading the fuzzy rule from said rule memory; generating a first adaptation degree of the input variable to express scalar quantity about a membership function designated by said read fuzzy rule when the input variable of the rule is a scalar quantity; generating a second adaptation degree of the input variable to express pattern data about a pattern category designated by said read fuzzy rule when the input variable is image pattern data; and applying a predetermined operation to one of said first and second adaptation degrees to generate a fuzzy inference output.

Claim 18

A fuzzy inference method according to claim 12, wherein said expressing step includes the further substep of inputting said image pattern data from an imaging device.

Claim 21

A method for generating an adaptation degree according to claim 14, wherein said previously storing step includes the further substep of inputting said image pattern data from an imaging device.

Claim 22

A pattern input type membership value generating method, comprising the steps of: inputting image pattern data; converting a category related to said image pattern data into standard pattern data about said category; computing a distance between said image pattern data and said standard pattern data; and converting said computed distance to an adaptation degree, said adaptation degree is a value "1" when the distance is zero and decreases as the distance increases.

Claim 23

A pattern input type membership value generating method, comprising the steps of: inputting image pattern data; converting a category related to said image pattern data into standard pattern data about said category; computing a distance between said image pattern data and said standard pattern data; and converting said computed distance to an adaptation degree, said adaptation degree is a value "1" when the distance is zero and decreases as the distance increases.

Claim 24

A membership value generating method according to claim 16, further comprising the steps of: generating a feature vector of standard pattern data; extracting a feature quantity of input image pattern data; generating a feature vector based on the extracted feature quantity; and computing a distance between the feature vector of the standard pattern data and the feature vector of input image pattern data.

Claim 25

A membership value generating method according to claim 16, further comprising the step of executing a pattern matching of said standard pattern data with said input image pattern data to compute said distance in accordance with a degree of inconsistency between said standard pattern data and said input image pattern data.

Claim 26

A pattern input type membership value generating method according to claim 16, wherein said image pattern data is input from an imaging device.