System and method for functional recognition of emitters

• US 5,884,294 A
• Filed: 04/18/1997
• Issued: 03/16/1999
• Est. Priority Date: 04/18/1997
• Status: Expired due to Term

• Current Assignee: Northrop Grumman Systems Corporation (Northrop Grumman Corporation)
• Original Assignee: Northrop Grumman Corporation
• Inventors: Kadar, Ivan, Schellhammer, Scott A.
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/844,685
• Time in Patent Office: 697 Days
• Field of Search: 706/10, 706/52, 706/49, 395/180, 395/182.09, 382/155, 382/156, 382/157, 382/158, 382/159
• US Class Current: 706/10
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06N 3/043 based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]

Abstract

The present invention relates to a radar emitter recognition system and method for classifying incoming emitters into their functional roles as disclosed by their electronic signatures and the context within which the radar transmissions are received. An electronic support system provides incoming data, including unknown attribute information, to a neural network which has been synthesized or trained to calculate a network solution indicative of an emitter'"'"'s classification within a range of attributes. Similarly the electronic support system provides incoming data, including unknown context data, to a fuzzy logic system that has been provided with possibility distributions to classify the emitter as originating from one of several strategic contexts. The resultant categorizations from the neural network and the fuzzy logic system are combined in a classifier to yield an improved recognition of the emitter under observation. This system and method of using a neural network and fuzzy logic is also applicable to other recognition problems in a number of fields.

Claims

1. A system for recognizing and classifying features of incoming data comprising:

input means for inputting synthesized or actual data representative of an attribute and a context of an emitter;

neural network means having an input to receive said incoming data representative of the attribute and providing an output network solution responsive to the emitter'"'"'s functional role;

logic means having an input to receive said incoming data representative of the context, one or more stored possibility distributions, a membership function calculation means, and one or more outputs indicative of the emitters grade of membership based on a correspondence between the possibility distributions and the context; and

means for classifying the emitter by calculating a probability distance measure based on the product of the network solution and the grade of membership.

2. A recognition system according to claim 1, wherein said neural network means comprises a synthesized neural network that is trained by extracting attributes from a training data base to thereby encode said neural network to classify the incoming data.

3. A recognition system according to claim 2, wherein said synthesized neural network contains a polynomial network means and a supervised abductive learning algorithm.

4. A recognition system according to claim 2, wherein said training data base comprises attributes based on a priori observations.

5. A recognition system according to claim 1, wherein the logic means includes means for extending conventional Boolean logic principles to the case of partial truth values between logical extremes of true and false.

6. A recognition system according to claim 1, wherein the logic means uses algorithms in combination with rules of inference to determine grades of membership of unknown emitters when compared to known emitters.

7. A recognition system according to claim 6, wherein a max-min operation is the rule of inference used in determining a correlation distance.

8. A recognition system according to claim 7 wherein the max-min correlation is used in combination with the output network solution and grade of membership to produce a probability distance metric.

9. A recognition system according to claim 1, wherein the logic means is a fuzzy logic system.

10. A recognition system according to claim 9, and further comprising means for updating neural network and fuzzy logic possibility distributions for adding new emitters and known emitters which have changed colors.

11. A recognition system according to claim 1, wherein said possibility distributions are obtained from a prior observations of known context and fitted into two normalized possibility distributions expressed by:

space="preserve" listing-type="equation">F.sub.P (x)=exp{-1/2 x-μ

P/σ

P!.sup.2 }

space="preserve" listing-type="equation">F.sub.B (x)=exp{-1/2 x-μ

B/σ

B!.sup.2 }.

12. A recognition system according to claim 1, wherein one or more of said outputs of said logic means indicates the emitter'"'"'s grade of membership as a color.

13. A recognition system according to claim 1, wherein one or more of said outputs of said logic means indicates the emitter'"'"'s grade of membership as a new emitter.

14. A method for recognizing and classifying features of incoming data, comprising the steps of:

inputting synthesized or actual data representative of an attribute and a context of an emitter;

applying the attribute data to a neural network which provides a network solution responsive to the emitter'"'"'s functional role;

applying the context data to a logic means having one or more stored possibility distributions for computing a grade of membership; and

calculating a probability distance measure based on the product of the network solution and the grade of membership.

15. A method for recognizing and classifying features of incoming data according to claim 14, wherein the step of applying the attribute data to a neural network includes a step of using a training data base based on synthesized or actual observations and encoding said network therefrom to thereafter classify the incoming data.

16. A method for recognizing and classifying features of incoming data according to claim 15, wherein the step of using includes a step of creating a training data base that includes the steps of:

a. creating an attribute knowledge base based on measurements of a set of emitter'"'"'s attributes including a transmitter'"'"'s center frequency, pulse repetition frequency, and sector coverage; and

b. sorting the sector coverage into a plurality of sets of differing angular measure with a constant standard deviation.

17. A method for recognizing and classifying features of incoming data according to claim 14, wherein the stored possibility distributions are stored in accordance with the steps of:

a. obtaining a prior observations of a set of emitter'"'"'s contexts including transmitter power and bearing;

b. sorting the bearing into sets of differing metric measurements creating a bearing context knowledge data base;

c. sorting the transmitter power into sets of differing power with a constant standard deviation creating a simulated power context knowledge base; and

d. using said simulated context knowledge bases to construct one or more normalized possibility distributions; and

e. storing the normalized possibility distributions for use by the logic means.