Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Electrical signal parameter measurement system
Patent
1997-02-12
1999-07-13
Wachsman, Hal Dodge
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Electrical signal parameter measurement system
702 70, 702189, 706900, G06F 1700
Patent
active
059240520
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method for detecting signals by means of fuzzy-logic classification. Such methods are required in various technical fields for controlling complex systems. In addition to system analysis, the control of a complex system also involves signal analysis for the purpose of being able to draw conclusions for controlled intervention from knowledge of the signal. It is frequently the came that only selected signal parameters are not always adequate but that it should be possible to evaluate the signal waveshape in as complex a way as possible.
2. Description of the Related Art
It is known from U.S. Pat. No. 5,000,188 to classify blood pressure curves in order to determine a person's age. For this purpose, the blood pressure curve of a specific person is examined and compared with representative curves of specific age groups by determining a measure of correlation between the curves, as a result of which the curve which is most similar to the current curve of the test subject is determined from the set of reference curves. The age of the text subject is then deduced from the age of the reference group. Further classification methods for curves are not known.
SUMMARY OF THE INVENTION
It is an object of the invention to provide a method for detecting signals by means of fuzzy-logic classification which satisfies the requirements mentioned at the beginning. This and other objects and advantages are achieved according to the invention by a method for detecting signals by means of fuzzy-logic classification wherein the range of signal amplitude values is subdivided into individual discrete threshold amplitude values; a membership function is defined for each discrete amplitude value in such a way that each has its respective maximum membership in the respective discrete amplitude value; at least one first reference pattern of a first signal to be detected is sampled equidistantly in time, a first reference signal amplitude being determined at each sampling instant; at each sampling instant, there is defined for at least each first reference signal amplitude a first assignment rule which specifies within which interval of two precisely designated directly neighboring threshold amplitude values the first reference amplitude is located at the respective sampling instant; a signal to be detected is sampled equidistantly in time in a fashion identical to the reference pattern, a detection signal amplitude being determined at each sampling instant; for the respective detection signal amplitudes, the membership values are determined in relation to the membership functions of the discrete amplitude values; the membership values are evaluated using the assignment rules as fuzzy-logic rules, and their strength is determined for the respective rules; the strengths of all the fuzzy-logic rules belonging to a respective reference pattern are added up and the arithmetic mean is formed therefrom as a measure of evaluation; the signal to be detected is detected as a first signal if the measure of evaluation is larger than 0.5. A signal catalog is recorded for this purpose and a signal to be identified in compared to it. This catalog can be recorded by a training phase or be produced width the cooperation of an expert. A suitable interval measure is used to determine significant differences. Consequently, only those signals which differ in accordance with this measure are recorded in the signal catalog. In accordance with the invention, a signal occurring is sampled equidistantly in time, and the amplitudes of the sampled values are compared with equidistant threshold values. The equidistant threshold values are assigned triangular membership functions which are equally distributed in the interval under consideration. The set of master signals which is to be identified is always assigned the same set of membership functions. For a selected master signal, membership values are determined for the sampling instant values from the membership functio
REFERENCES:
patent: 5000188 (1991-03-01), Kojima
patent: 5245695 (1993-09-01), Barehore
patent: 5287432 (1994-02-01), Tomitaka
patent: 5491775 (1996-02-01), Madau et al.
Holger Franke, Chemnitz, "parallel-FUCS: Einsatzsystem zur On-line-und Real-time-Fuzzy-Klassifikation", 8129 AT Automatisierungstechnik 42(1994) Mar., No. 3, Germany, pp. 118-123. ( No English Translation).
R. Marin, "On knowledge-based fuzzy classifiers: A medical case study", 8355 Fuzzy Sets and Systems 44(1991) Dec. 16, No. 3, Amsterdam, NE, pp. 421-430.
T. Tilli, "Mustererkennung mit Fuzzy-Logik", 1993, pp. 211-257. (No English Translation) (No date with month).
Siemens Aktiengesellschaft
Wachsman Hal Dodge
LandOfFree
Method for detecting signals by means of fuzzy-logic classificat does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for detecting signals by means of fuzzy-logic classificat, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for detecting signals by means of fuzzy-logic classificat will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2288166