Data processing: measuring – calibrating – or testing – Measurement system – Statistical measurement
Reexamination Certificate
2000-05-16
2002-11-19
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system
Statistical measurement
C702S022000, C702S085000, C702S189000
Reexamination Certificate
active
06484122
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to the field of signal processing. More specifically, the present invention relates to signal processing a characteristic signal of a subject.
BACKGROUND OF THE INVENTION
In industrial automation, signal processing is used to classify an object being manufactured or processed based on a characteristic of the object. For example, an apple might be classified by a weight sensor configured to sense the weight of the apple. If the weight is greater than a predetermined weight, the apple is identified as “good”, and, if not, the apple is identified as “bad”.
However, the object can also be classified by other signals. For example, the apple might also be classified by acquiring a color digital image of the apple. If the apple is darker than a predetermined gray scale, or if the apple lacks sufficient red color, the apple is identified as “bad”. The challenge is to determine which characteristic (e.g., weight, color, gray scale, etc.) best classifies the objects into the desired classifications, so that the best characteristic can be used during production to automatically classify objects.
A standard method for evaluating the classification of objects has been implemented which assumes a bimodal distribution of the measured characteristic, the distributions assumed to be Gaussian. For example, referring to
FIG. 1
, this standard method generates a histogram
9
of the frequency of occurrence of different values of the characteristic. The x-axis represents the values of the characteristic (e.g., weight, color, etc.) and the y-axis represents the frequency of objects having that characteristic. A first mode
11
includes objects in a first class (e.g., “bad” objects) and a second mode
13
includes objects in a second class (e.g., “good” objects). According to this method, the mean values
17
,
15
of each mode are identified, the variances of mean values
17
,
15
are determined, and the distance
19
between mean values
17
and
15
is determined. The smaller the variances and the greater the interval between mean values
17
,
15
, the greater is the quality of the characteristic for classification of this object.
One drawback of this method is that characteristic distributions frequently are neither bimodal nor Gaussian and, thus, are incorrectly evaluated by this prior method. With reference to
FIG. 2
, a frequency distribution
21
of another characteristic is shown, in which mode
23
is not Gaussian. Further, mode
23
includes objects in a first class, mode
24
includes objects in a second class, and mode
26
includes additional objects in the first class. An example of such a distribution might be one in which the characteristic is the length of a wooden dowel, wherein “good” dowels must have a length within a certain tolerance. Thus, “bad” dowels have lengths greater than (mode
26
) and less than (mode
23
) “good” dowels (mode
24
). Prior methods will not adequately evaluate the suitability of this characteristic for classification purposes, since the distribution in
FIG. 2
is not Gaussian and not bimodal.
Accordingly, there is a need for a system and method for evaluating the suitability of characteristics for classification. There is further a need for such a system and method which is applicable to non-Gaussian distributions. Further still, there is a need for such a system and method which is applicable to non-bimodal distributions. There is also a need for such a system and method which is robust against noise.
SUMMARY OF THE INVENTION
According to an exemplary embodiment, a method of evaluating a characteristic for suitability in classification of subjects based on subject data is provided. The subject data includes characteristic data and class data. The method includes arranging the subject data based on the characteristic data, and identifying the number of class changes from one class to another class in the arranged subject data. The number of class changes represents the suitability of the characteristic for classification of the subjects.
According to an alternative embodiment, a method of evaluating a characteristic for suitability in classification of subjects based on subject data is provided. The subject data includes characteristic data and class data. The method includes arranging the subject data based on the characteristic data, identifying consecutive subject data having a class change, and measuring the interval between the two consecutive subject data. The interval between class changes represents the suitability of the characteristic for classification of the subject.
According to yet another alternative embodiment, a system for evaluating a characteristic for suitability in classification of subjects is provided. The system includes sensing means for acquiring characteristic data from a plurality of subjects and classification means for classifying each subject with one of a first class and a second class. The system further includes means for arranging the subject data based on the characteristic data and identifying the number of class changes from one class to another class in the arranged subject data. The number of class changes represents the suitability of the characteristic for classification of the subjects.
REFERENCES:
patent: 5555406 (1996-09-01), Nozawa
patent: 5715327 (1998-02-01), Wilhelm et al.
patent: 5987398 (1999-11-01), Halverson et al.
patent: 6253159 (2001-06-01), Bett et al.
patent: 6269313 (2001-07-01), Givens et al.
Wolfgang Abmayr, Einführung in die digitale Bildverarbeitung, 1994, pp. 223-233 and 272-275.
Foley & Lardner
Hoff Marc S.
Siemens Aktiengesellschaft
Suarez Felix
LandOfFree
System and method for evaluating characteristics for... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for evaluating characteristics for..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for evaluating characteristics for... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2993139