Apparatus and method for automatically creating membership functions

• US 5,402,525 A
• Filed: 08/06/1993
• Issued: 03/28/1995
• Est. Priority Date: 08/26/1992
• Status: Expired due to Term

• Current Assignee: U Krowel LLC (Intellectual Ventures LLC)
• Original Assignee: Omron Corporation
• Inventors: Ishida, Tsutomu, Okumura, Akira
• Primary Examiner(s): Knepper, David D.
• Application Number:
  US08/102,712
• Time in Patent Office: 599 Days
• Field of Search: 395/3, 395/51, 395/61, 395/900, 395/76
• US Class Current: 706/60
• CPC Class Codes: G06N 7/023 Learning or tuning the parameters of a fuzzy system
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

Membership functions are capable of being created with ease even by a person who is not an expert or skilled individual. Data Ai˜Bi sensed by sensors 11˜13 and synthesized data Mi are stored in a memory device 31 in correlation with identification codes Di of objects. The results of comparing the "tastiness" of an object Di with that of an object Dj are entered from an input unit 15 and stored in the memory device 31. An evaluating unit 32 arrays the results of comparison in accordance with inequality symbols and assigns evaluation values to these results in the order of the array. A histogram creating unit 33 generates a histogram by plotting each item of input data Ai˜Mi along a horizontal axis and plotting the corresponding evaluation values along a vertical axis. Whether or not a function obtained by connecting the peaks of the histogram is appropriate as a basic function for creation of a membership function is judged by a judging unit 34 from the viewpoint of whether the function has a unique peak or a variance about the peak that is small. A function judged to be appropriate is normalized by a generating unit 35, thereby generating a membership function.

Claims

1. An apparatus for automatically creating membership functions, comprising:

first input means for entering objective numerical-value data regarding an object along with identification codes of this data;

first memory means for storing, in correlation with the identification codes, the numerical-value data entered by said first input means;

second input means for entering results of subjective judgment regarding the object, these results being entered as comparative data represented using the identification codes of the numerical-value data entered by said first input means;

second memory means for storing the comparative data entered by said second input means;

evaluating means for assigning evaluation values in correspondence with the identification codes in accordance with a ranking based upon the comparative data stored in said second memory means;

means for creating a histogram by arraying the numerical-value data, which has been stored in said first memory means, on a horizontal axis in accordance with the numerical values, and plotting, along a vertical axis, the evaluation values assigned to correspond to the identification codes associated with the numerical-value data;

judging means for judging whether the histogram created by said histogram creating means is appropriate as a basis for a membership function; and

membership-function generating means for generating a membership function based upon a histogram judged to be appropriate by said judging means.

2. The apparatus according to claim 1, wherein said first input means is a plurality of sensors for performing measurements regarding a plurality of attributes of an object, and said histogram generating means creates a histogram for each item of data obtained from said plurality of sensors.

3. The apparatus according to claim 1, wherein a plurality of said first input means are provided, a plurality of the second input means are provided, in correspondence with respective ones of said plurality of first input means, for entering the results of subjective judgment relating to the numerical-value data entered by said first input means, and said histogram creating means creates a histogram based upon mutually corresponding input data from said first input means and comparative data from said second input means.

4. The apparatus according to claim 1, wherein a plurality of said first input means are provided, means are provided for creating synthesized numerical-value data based upon at least two types of numerical-input value from said plurality of first input means, and said histogram creating means creates a histogram using the synthesized numerical-value data.

5. A method of automatically creating membership functions comprising the steps of:

entering objective numerical-value data regarding an object along with identification codes of this data;

storing, in correlation with the identification codes, the entered numerical-value data;

entering results of subjective judgment regarding the object, these results being entered as comparative data represented using the identification codes of the entered numerical-value data;

storing the entered comparative data;

assigning evaluation values in correspondence with the identification codes in accordance with a ranking based upon the comparative data;

creating a histogram by arraying the numerical-value data on a horizontal axis in accordance with the numerical values and plotting, along a vertical axis, the evaluation values assigned to correspond to the identification codes associated with the numerical-value data;

judging whether the created histogram is appropriate as a basis for a membership function; and

generating a membership function based upon a histogram judged to be appropriate.