Radio direction finding system for narrowband multiple signals

• US 5,768,477 A
• Filed: 09/10/1996
• Issued: 06/16/1998
• Est. Priority Date: 09/10/1996
• Status: Expired due to Term

• Current Assignee: Southwest Research Institute
• Original Assignee: Southwest Research Institute
• Inventors: Johnson, Richard L., Montville, William J.
• Primary Examiner(s): Davis, George B.
• Application Number:
  US08/711,040
• Time in Patent Office: 644 Days
• Field of Search: 395/22, 395/23
• US Class Current: 706/24
• CPC Class Codes: G01S 3/74 Multi-channel systems specially adapted for direction-finding, i.e. having a single antenna system capable of giving simultaneous indications of the directions of different signals systems in which the directions of different signals are determined sequentially and displayed simultaneously G01S3/04, G01S3/14

Abstract

A method of training a neural network to determine directions of arrival when there are multiple input signals within a narrow frequency band. For a single input signal having a given angle of arrival, an antenna array is used to sample signal values, from which spatial covariance matrix values are obtained. Each value is applied at an input node of the neural network. The neural network is adjusted so that an output node associated with that signal'"'"'s angle of arrival fires in response to these inputs. This process is repeated for different angles of arrival, such that for L different angles of arrival, L output nodes are trained with L different training sets of data. Once trained, the neural network can be used in a direction finding system that detects whatever number of signals are present in a received wavefield having any number of signals.

Claims

1. A method of using a single training signal to train a neural network to determine the number of signals in a received wavefield, and the azimuth angle of arrival of each, comprising the steps of:

receiving, with an array of antennas, a training signal from a known azimuth direction;

converting the output of each said antenna to a digital value representing a complex signal value;

repeating said receiving and converting steps at a number of time intervals, thereby obtaining sample vectors, one said vector at each time interval;

calculating, from said vectors, a matrix of spatial covariance values, said values comprising a training set;

applying each value of said training set to an input node of said neural network, said neural network having an number of output nodes, each associated with a different azimuthal direction;

adjusting said neural network so that an output node of said neural network associated with said known azimuth direction of said signal fires in response to said applying step; and

repeating all of the above steps with said training signal from a number of known azimuth directions, each of said known azimuth directions associated with a different output of said neural network.

2. The method of claim 1, wherein said number of azimuthal directions are even increments of a range of azimuth directions.

3. The method of claim 1, wherein said increments are less than the beamwidth of said antenna array.

4. The method of claim 1, wherein said neural network has additional output nodes, each associated with a different altitude, and further comprising the step of repeating the method with said training signal at a number of different altitudes such that a different output node of said neural network fires in response to said training signal at each said different altitude.

5. A method of using a single training signal to train a neural network to determine the number of signals in a received wavefield in a received wavefield and the azimuth angle of arrival of each, comprising the steps of:

receiving, with a linear array of antennas, having substantially equal spacing, a training signal from a known azimuth direction;

converting the output of each said antenna to a digital value representing a complex signal value;

repeating said receiving and converting steps at a number of time intervals, thereby obtaining sample vectors, one said vector at each time interval;

calculating, from said vectors, a row or column of a matrix of spatial covariance values;

applying each value of said row or column to an input node of said neural network, said neural network having an number of output nodes, each associated with a different azimuth direction;

adjusting said neural network so that an output node of said neural network associated with said known azimuth direction of said signal fires in response to said applying step; and

repeating all of the above steps with said training signal for a number of known azimuth directions, each said known azimuth direction associated with an output of said neural network.

6. The method of claim 5, wherein said number of azimuth directions are even increments of a range of azimuth directions.

7. The method of claim 5, wherein said increments are less than the beamwidth of said antenna array.

8. The method of claim 5, wherein said neural network has additional output nodes, each associated with a different altitude, and further comprising the step of repeating the method with said training signal at a number of different altitudes such that a different output node of said neural network fires in response to said training signal at each said different altitude.

9. A method of using a single training signal to train a neural network to the number of signals in a received wavefield, and the azimuth angle of arrival of each, said signals having an expected phase, comprising the steps of:

receiving, with an array of antennas, a training signal from a known azimuth direction having substantially the same phase as said expected phase;

converting the output of each said antenna to a digital value;

repeating said receiving and converting steps at a number of time intervals, thereby obtaining sample vectors, one said sample vector at each time interval;

calculating, from said sample vectors, average amplitude and phase values;

applying each said average amplitude and phase value to an input node of said neural network, said neural network having an number of output nodes, each associated with a different azimuthal direction;

adjusting said neural network so that an output node of said neural network associated with said known azimuth direction of said signal fires in response to said applying step; and

repeating all of the above steps with said training signal for a number of known azimuth directions, each said known azimuth direction associated with an output of said neural network.

10. The method of claim 9, wherein said number of azimuth directions are even increments of a range of azimuth directions.

11. The method of claim 9, wherein said increments are less than the beamwidth of said antenna array.

12. The method of claim 9, wherein said neural network has additional output nodes, each associated with a different altitude, and further comprising the step of repeating the method with said training signal at a number of different altitudes such that a different output node of said neural network fires in response to said training signal at each said different altitude.

13. A method of using a single training signal to train a neural network to determine the number of signals in a received wavefield, and the azimuth angle of arrival of each, said signals being substantially coherent, comprising the steps of:

receiving, with an array of antennas, a training signal from a known azimuth direction;

converting the output of each said antenna to a digital value representing a complex signal value;

repeating said receiving and converting steps at a number of time intervals, thereby obtaining sample vectors, one said vector at each time interval;

calculating, from said vectors, a matrix of spatial covariance values, said values comprising a training set;

dividing said matrix into submatrices and averaging corresponding values of each of said submatrices, thereby obtaining a smoothed training set;

applying each value of said smoothed training set to an input node of said neural network, said neural network having an number of output nodes, each associated with a different azimuthal direction;

adjusting said neural network so that an output node of said neural network associated with said known azimuth direction of said signal fires in response to said applying step; and

repeating all of the above steps with said training signals from a number of known azimuth directions, each of said known azimuth directions associated with a different output of said neural network.

14. The method of claim 13, wherein said number of azimuth directions are even increments of a range of azimuth directions.

15. The method of claim 13, wherein said increments are less than the beamwidth of said antenna array.

16. The method of claim 13, wherein said neural network has additional output nodes, each associated with a different altitude, and further comprising the step of repeating the method with said training signal at a number of different altitudes such that a different output node of said neural network fires in response to said training signal at each said different altitude.

17. A method of using a neural network to determine the number of signals in a received wavefield, comprising the steps of:

receiving said wavefield with an array of antennas;

converting the output of each said antenna to a digital value;

applying each said digital value to an input node of said neural network, said neural network having an number of output nodes, each associated with a different azimuthal direction, and said neural network having been trained with a single training signal varied over time from different directions of arrival; and

estimating the number of signals in said wavefield by detecting how many of said output nodes of said neural network respond to said applying step.

18. The method of claim 17, further comprising the step of determining the azimuth angle of arrival of each of said signals by detecting which of said output nodes respond to said applying step.