Apparatus for performing non-linear signal classification in a communications system

• US 6,240,282 B1
• Filed: 07/13/1998
• Issued: 05/29/2001
• Est. Priority Date: 07/13/1998
• Status: Expired due to Term

• Current Assignee: General Dynamics C4 Systems Incorporated (General Dynamics Corporation)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Kleider, John Eric, Bergstrom, Chad Scott
• Primary Examiner(s): CUMMING, WILLIAM D
• Application Number:
  US09/114,321
• Time in Patent Office: 1,051 Days
• Field of Search: 455/226.1, 375/285
• US Class Current: 455/226.1
• CPC Class Codes: H04B 17/20 of receivers
• CPC Class Codes: H04L 27/0012 arrangements for identifying the type of modulation

Abstract

A non-linear signal classifier (26) includes a polynomial expansion unit for expanding signal feature vectors determined by a feature extraction unit (25) for a received signal. The expanded signal feature vectors are each combined with a plurality of signal classification models that are stored in a model memory (76). The signal classification models are each associated with a particular signal type that is recognized by the non-linear signal classifier. A scoring unit (72) generates a score for each of the signal classification models based on the result of the combination. The scores are analyzed by a selection unit (74) which determines which of the signal classification models (i.e., which of the signal types) most likely represents the received signal. Training equipment (60) is also provided for training the non-linear signal classifier (26) to recognize new signal types. In one embodiment, the training equipment (60) is capable of adding a new signal classification model to the model memory (76) without modifying other models stored therein.

Claims

1. A receiver for use in a communications system, said receiver comprising:

a signal receptor for receiving a communications signal from a communications channel;

a feature extraction unit for extracting at least one feature from said communications signal, said at least one feature forming a feature set;

a non-linear classifier for classifying said communications signal according to signal type, said non-linear classifier including a polynomial expansion unit for performing a polynomial expansion on said at least one feature to generate an expanded feature set, said non-linear classifier using said expanded feature set to classify said communications signal;

a signal processor for processing said communications signal based on signal type determined by said non-linear classifier;

said non-linear classifier includes a model memory for storing a plurality of signal classification models, each signal classification model corresponding to a particular signal type in a signal type set;

said non-linear classifier includes a combination unit for combining said expanded feature set with each of said plurality of signal classification models to produce at least one combination value set for each of said plurality of signal classification models;

said expanded feature set includes at least one expanded feature vector;

said signal classification models each include at least one signal classification vector; and

said combination unit includes means for finding a vector dot product between an expanded feature vector and a signal classification vector.

2. The receiver, as claimed in claim 1, wherein:

said non-linear classifier includes a scoring unit for determining a score for each of said signal classification models based on said at least one combination value set associated therewith.

3. The receiver, as claimed in claim 2, wherein:

said scoring unit includes means for determining an average value for said at least one combination value set.

4. The receiver, as claimed in claim 2, wherein:

said non-linear classifier includes a selection unit for selecting a signal classification model having a best score.

5. The receiver, as claimed in claim 1, further comprising:

a signal preconditioning unit, located between said signal receptor and said feature extraction unit, for conditioning said communications signal in a manner that facilitates extraction of features.

6. The receiver, as claimed in claim 5, wherein:

said signal preconditioning unit includes means for improving a signal-to-noise ratio (SNR) of said communications signal.

7. The receiver, as claimed in claim 6, wherein:

said means for improving includes means for estimating a bandwidth of a component of said communications signal.

8. The receiver, as claimed in claim 7, wherein:

said means for estimating a bandwidth includes means for converting said communications signal to a frequency domain representation including frequency domain coefficients.

9. The receiver, as claimed in claim 8, wherein:

said means for estimating a bandwidth includes means for determining a noise floor for said signal using said frequency domain coefficients.

10. The receiver, as claimed in claim 7, wherein:

said means for improving includes means for filtering out portions of said communications signal outside said estimated bandwidth.

11. The receiver, as claimed in claim 6, wherein:

said means for improving includes a phase domain filter for filtering de-aliased phase components of the communications signal.

12. The receiver, as claimed in claim 5, wherein:

said signal preconditioning unit includes means for separating a desired component of said communications signal from an interference component of said communications signal.

13. The receiver, as claimed in claim 1, wherein:

said non-linear classifier operates in a multi-resolutional mode wherein classification is performed in a plurality of classification levels.

14. The receiver, as claimed in claim 1, wherein:

said feature extraction unit operates in a multi-resolutional mode wherein feature extraction is performed in a plurality of feature classification levels.

15. The receiver, as claimed in claim 1, wherein:

said feature extraction unit and said non-linear classifier operate in a multi-resolutional mode.

16. The receiver, as claimed in claim 1, wherein:

said feature extraction u nit includes a second non-linear classifier including a second polynomial expansion unit, wherein said second non-linear classifier is used to determine features to be extracted.

17. The receiver, as claimed in claim 1, wherein:

said signal processor includes a parameter extraction unit for extracting parameter values from said communications signal, wherein the parameters for which values are extracted depend upon the signal type determined by the non-linear classifier.

18. The receiver, as claimed in claim 1, wherein:

said signal processor includes a differential decoder for removing differential encoding (DE) from said communications signal, said differential decoder being responsive to at least one control signal that indicates a form of differential encoding present in said communications signal, wherein said at least one control signal is generated using said signal type determined by said non-linear classifier.

19. The receiver, as claimed in claim 1, wherein:

said signal processor includes a demodulator for demodulating said communications signal, said demodulator being responsive to at least one control signal that indicates a form of modulation present in said communications signal, wherein said at least one control signal is generated using said signal type determined by said non-linear classifier.

20. The receiver, as claimed in claim 1, wherein:

said signal processor includes means for determining a form of spread spectrum modulation that is present in said communications signal using said signal type determined by said non-linear classifier, said means for determining including means for generating at least one control signal indicative of said form of spread spectrum modulation.

21. The receiver, as claimed in claim 20, wherein:

said means for determining a form of spread spectrum modulation includes means for determining a frequency hopping code and means for determining a pseudo noise code.

22. The receiver, as claimed in claim 20, wherein:

said signal processor includes means for removing spread spectrum modulation from said communications signal to generate an output signal, said means for removing being responsive to said at least one control signal generated by said means for generating.

23. The receiver, as claimed in claim 22, wherein:

said signal processor includes feedback means for use in transferring said output signal back to said non-linear classifier for further classification.

24. A receiver for use in a communications system, comprising:

a signal receptor for receiving a communications signal from a communications channel;

a feature extraction unit for extracting at least one feature vector from said communications signal, each of said at least one feature vector corresponding to a different feature of said communications signal;

a non-linear classifier, coupled to said feature extraction unit, for classifying said communications signal according to signal type, said non-linear classifier including;

a polynomial expansion unit for expanding said at least one feature vector by generating cross-products between elements of said at least one feature vector to generate at least one expanded feature vector;

a model memory storing a plurality of signal classification models for use in classifying said communications signal, wherein each of said signal classification models corresponds to a predetermined signal type;

a combination unit for combining said at least one expanded feature vector with each of at least two signal classification models in said model memory to produce at least two combination value sets, wherein each combination value set corresponds to a different signal classification model; and

means for comparing said at least two combination value sets to determine a most likely signal type for said communications signal;

a signal processor for processing said communications signal using said classification signal;

each of said plurality of signal classification models includes at least one classification vector; and

said combination unit includes means for calculating a vector dot product between an expanded feature vector and a classification vector within a first signal classification model.

25. The receiver, as claimed in claim 24, wherein:

said at least one expanded feature vector includes multiple expanded feature vectors; and

said combination unit includes means for calculating a vector dot product between each of said multiple expanded feature vectors and a classification vector within said first classification model to generate a first combination value set, wherein said first combination value set includes as elements thereof the results of the vector dot products.

26. The receiver, as claimed in claim 24, wherein said means for comparing comprises:

a scoring unit for determining a score for each of said at least two signal classification models using said at least two combination value sets; and

a selection unit for selecting one of said at least two signal classification models based on scores determined by said scoring unit, said selection unit outputting a classification signal that is indicative of a signal type associated with said selected signal classification model.

27. The receiver, as claimed in claim 26, wherein:

each of said at least two combination value sets comprises a combination vector; and

said scoring unit calculates an average value for elements in each combination vector to determine a score for a corresponding signal classification model.

28. The receiver, as claimed in claim 27, wherein:

said selection unit selects a signal classification model having a highest average value.

29. The receiver, as claimed in claim 24, wherein:

said means for comparing outputs a signal identifying a number of different signal types and indicating a probability that each of said number of different signal types is the signal type of the communications signal being tested.

30. An apparatus for performing signal classification in a communications system, said apparatus comprising:

an input for receiving at least one feature vector associated with a communications signal;

a polynomial expansion unit, coupled to said input, for performing a polynomial expansion on said at least one feature vector, said polynomial expansion unit outputting at least one expanded feature vector; and

a testing unit, coupled to said polynomial expansion unit, for determining a signal type of said communications signal, said testing unit comprising;

a model memory for storing a plurality of signal classification models, wherein each of said plurality of signal classification models corresponds to a predetermined signal type that is recognized by said apparatus;

a combination unit, coupled to said model memory and said polynomial expansion unit, for combining said at least one expanded feature vector with at least two of said plurality of signal classification models to generate at least two combination value sets, wherein each combination value set corresponds to a different signal classification model;

means for comparing said at least two combination value sets to determine a signal type of said communications signal;

a training unit coupled to said polynomial expansion unit for training said apparatus to recognize a new signal type, said training unit including;

means for receiving a first expanded feature set from said polynomial expansion unit for a signal having said new signal type; and

means for generating a new signal classification model for said new signal type using said first expanded feature set; and

means for storing said new signal classification model in said model memory without modifying other signal classification models in said model memory, wherein said other signal classification models in said model memory are still capable of use in identifying signal type even though said other signal classification models have not been modified.

31. The apparatus, as claimed in claim 30, wherein:

said polynomial expansion unit calculates cross products of vector elements within said at least one feature vector up to a desired order n.

32. The apparatus, as claimed in claim 30, further comprising:

a mode switch coupled to said polynomial expansion unit, said testing unit, and said training unit for directing expanded feature sets from said polynomial expansion unit to either said testing unit or said training unit based on a control signal received from a controller.

33. The apparatus, as claimed in claim 32, further comprising:

out-of-class logic, coupled to an output of said testing logic, for determining when a first communications signal does not fall within a recognized signal type, said out-of-class logic including means for instructing the controller to train the apparatus to recognize a signal type of said first communications signal.

34. In a communications system that includes a signal classifier for classifying a receive signal according to a plurality of recognized signal types, an apparatus for use in training the signal classifier to recognize a new signal type, said apparatus comprising:

a model memory for storing signal classification models corresponding to the plurality of recognized signal types, said signal classification models for use by the signal classifier in classifying the receive signal;

a model generation unit for generating a new signal classification model, corresponding to the new signal type, for storage in said model memory, wherein said new signal classification model can be generated and stored in said model memory without modifying signal classification models previously stored in said model memory, wherein said signal classification models previously stored in said model memory are still capable of use in classifying the receive signal after said new signal classification model has been stored even though they have not been modified;

said model memory includes signal classification models that are organized based on corresponding values for two or more signal characteristics;

said model memory includes signal classification models arranged in rows and columns, wherein said rows each correspond to a different value of a first signal characteristic and said columns each correspond to a different value of a second signal characteristic; and

said rows each correspond to a different signal class and said columns each correspond to a different signal-to-noise ratio (SNR).

35. The apparatus, as claimed in claim 34, wherein:

said model generation unit includes an input for receiving a feature set corresponding to said new signal type.

36. The apparatus, as claimed in claim 35, wherein:

said feature set includes at least one feature vector that has been expanded using a polynomial expansion.

37. The apparatus, as claimed in claim 34, wherein:

said signal classification models are organized in levels, wherein each level corresponds to a particular signal characteristic.

38. The apparatus, as claimed in claim 34, wherein:

said signal classification models are organized to support multi-resolutional signal classification.

39. The apparatus, as claimed in claim 34, further comprising:

a feature extraction unit for extracting at least one feature vector from an input signal having the new signal type, said at least one feature vector describing at least one feature of said input signal; and

a polynomial expansion unit for performing a polynomial expansion on said at least one feature vector by calculating cross products of elements of said at least one feature vector and adding said cross products to said at least one feature vector as further elements of said at least one feature vector to produce at least one expanded feature vector;

wherein said polynomial expansion unit is coupled to said model generation unit for delivering said at least one expanded feature vector to said model generation unit.

40. The apparatus, as claimed in claim 39, wherein:

said at least one expanded feature vector includes a plurality of expanded feature vectors; and

said model generation unit includes means for determining a high order correlation vector for the new signal type by adding together said plurality of expanded feature vectors using vector addition.

41. The apparatus, as claimed in claim 40, wherein:

said model generation unit includes means for determining a high order correlation sum vector for the plurality of recognized signal types by adding together high order correlation vectors associated with each of the plurality of recognized signal types including the high order correlation vector associated with the new signal type.

42. The apparatus, as claimed in claim 41, wherein:

said model generation unit includes means for determining a scaled high order correlation vector for the new signal type using said high order correlation sum vector.