Method and system for soft-weighting a reiterative adaptive signal processor

• US 8,082,286 B1
• Filed: 03/26/2007
• Issued: 12/20/2011
• Est. Priority Date: 04/22/2002
• Status: Expired due to Fees

• Current Assignee: Leidos, Inc. (Leidos Holdings, Inc.), Caliper Life Sciences Incorporated (Revvity, Inc.)
• Original Assignee: Science Applications International Corporation
• Inventors: Picciolo, Michael Lee, Gerlach, Karl Robert
• Primary Examiner(s): Bullock, Jr., Lewis
• Assistant Examiner(s): Hughes, Kevin G
• Application Number:
  US11/727,422
• Time in Patent Office: 1,730 Days
• Field of Search: 708/322
• US Class Current: 708/322
• CPC Class Codes: G01S 7/292 Extracting wanted echo-signals

Abstract

This invention is an improvement upon the basic adaptive signal processors, the Multi-Stage Wiener Filter and the cascaded canceller. The invention combines the concepts of soft weighting the adaptive weights of either type of processor disclosed herein with reiteratively processing the outputs by returning them to the input of the chosen adaptive signal processor. The combination of these functions improves the Signal to Interference plus Noise Ratio (SINR), the Probability of Detection (Pd), and/or the Bit Error Rate (BER). The invention improves statistical convergence of these types of metrics such that fewer training data samples are needed to achieve a particular satisfactory value of these metrics than would occur using traditional computational adaptive signal processors.

Claims

1. A method of training an adaptive signal processor for filtering interference and noise from an input signal received by the adaptive signal processor, which comprises a plurality of building blocks, each building block comprising an adaptive weight value, the method comprising:

i) receiving, by the adaptive signal processor, an input signal comprising a plurality of measurements, z and x, in an interference plus noise data environment;

ii) applying, by each of the plurality of building blocks of the adaptive signal processor, a soft weight complex scalar value, w_sw, to each of the plurality of measurements, the soft weight value multiplied with the adaptive weight value, w_g, thereby modifying the adaptive weight value;

iii) outputting, by the adaptive signal processor, a main channel column output, y, comprising the weighted plurality of measurements in accordance with the equation y=z−

(w_sw*w_g)*x;

iv) squaring, by the adaptive signal processor, the main channel column output of the adaptive signal processor for each of the plurality of measurements;

v) determining, by the adaptive signal processor, a representative average value of the squared values; and

vi) performing each of steps i) through v), by the adaptive signal processor, the performing step comprising;

verifying, when the representative average value is within a pre-determined signal to power plus noise ratio [SINR]; and

upon a positive verification that the representative average value is within a pre-determined signal to power plus noise ratio [SINR], counting the number of iterations of steps i) through v), whereupon, when the count is less than 2N training samples, N being the number of degrees of freedom of the adaptive signal processor, the performing step is concluded and the adaptive signal processor is now trained to filter between interference and noise of an input signal.

2. The method according to claim 1 wherein applying the soft weight value includes determining the soft weight value.

3. The method according to claim 2 wherein determining the soft weight value further comprises the steps of:

i) selecting a value;

ii) generating a data set of training values;

iii) providing an adaptive signal processor simulator;

iv) putting the training values into the simulator,

v) inputting the selected value into the simulator;

vi) measuring a signal power output value of the simulator;

vii) varying the selected value; and

viii) repeating steps i)-vii) until the measured difference in signal power output value between each preceding iteration is about 1 dB.

4. The method according to claim 3, wherein varying the selected value further comprises choosing the selected value from the range 0 to 2, excluding a value of 1.

5. The method according to claim 4 wherein the selected value is the soft weight value.

6. The method according to claim 3 wherein generating a data set of training values further comprises producing a first data matrix of size defined by a number of degrees of freedom [N] of a sensor input array and twice a minimum expected effective rank, [R_min], corresponding to a simulated interference plus noise environment.

7. The method according to claim 3 wherein generating a data set of training values further comprises producing a second data matrix of size defined by a number of degrees of freedom [N] of a sensor input array and twice a maximum expected effective rank, [R_min], corresponding to a simulated interference plus noise environment.

8. Tire method according to claim 3 wherein the soft weight value is selected such that the pre-determined average value is a signal to power plus noise ratio [SINR] within 3 dB from optimal normalized SINR of 0 dB, the soft weight value selected in less than 2N training samples, where N is the number of degrees of freedom.

9. The method according to claim 3 wherein the simulator simulates a cascaded canceller.

10. The method according to claim 9 wherein the cascaded canceller comprises a pseudo-median cascaded canceller.

11. The method according to claim 10 wherein at least one adaptive weight of the pseudo-median cascaded canceller is multiplied by the selected value.

12. The method according to claim 9 wherein the cascaded canceller comprises a Gram-Schmidt cascaded canceller algorithm.

13. The method according to claim 12 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and at least one Gram-Schmidt weight value of the Gram-Schmidt cascaded canceller is multiplied by the selected value.

14. The method according to claim 9 wherein the cascaded canceller comprises a generic cascaded canceller algorithm.

15. The method according to claim 14 wherein at least one adaptive weight value of the generic cascaded canceller is multiplied by the selected value.

16. The method according to claim 9 wherein the cascaded canceller comprises a multi-stage median cascaded canceller algorithm.

17. The method according to claim 16 wherein at least one adaptive weight value of the multi-stage median cascaded canceller is multiplied by the selected value.

18. The method according to claim 3 wherein the simulator simulates a Multi-Stage Weiner Filter algorithm.

19. The method according to claim 18 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages is selected to be equal to an effective rank of an interference environment.

20. The method according to claim 18 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages is selected to be greater than a minimum effective rank of an interference environment.

21. The method according to claim 18 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages required for optimal convergence is less than the number of degrees of freedom.

22. The method according to claim 18 wherein the one or more adaptive weight values is a Weiner weight value, and at least one Weiner weight value is multiplied by the selected value.

23. The method according to claim 18 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and at least one Gram-Schmidt weight value is applied in the synthesis stage.

24. The method according to claim 18 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and at least one Gram-Schmidt weight value is multiplied by the selected value.

25. The method according to claim 3 wherein the soft weight value is increased until the measured signal power output value reaches an unacceptable limit.

26. A system for reiteratively applying a soft-weight value, comprising:

a plurality of sensor inputs for measuring signals within an interference plus noise environment with a variable effective rank;

a filter comprising an adaptive signal processor;

a controller to perform the following steps;

input the measured signals, x and z, into the filter;

determining y, in accordance with the following equation;

y=z−

(w_sw*w_g)*x, wherein x and z are input signals, w_g is conjugated adaptive weight value, w_sw is representative of the soft weight complex scalar value

applying a soft weight value w_sw to the filter, the soft weight value modifying one or more existing adaptive weight values w_g associated with the filter;

outputting, y of the adaptive signal processor in accordance with the equation y=z−

(w_sw*w_g)*x;

a memory means, wherein the controller retrieves from the memory means a stored number of iterations, such that the operations of the controller are performed a number of times as specified by the number of iterations; and

an output means.

27. The system according to claim 26 wherein a plurality of degrees of freedom provided by the sensor inputs provides the measured target signals.

28. The system according to claim 26 wherein the memory means has stored upon it a number of iterations.

29. The system according to claim 28 wherein the controller retrieves from the memory means a stored number of iterations, such that the measured signals are input into the filter a number of times as specified by the number of iterations.

30. The system according to claim 26 wherein the adaptive signal processor is a Gram-Schmidt Cascaded Canceller.

31. The system according to claim 30 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and the controller applies the soft weight value by multiplying at least one Gram-Schmidt weight value of the Gram-Schmidt cascade canceller by the soft weight value.

32. The system according to claim 26 wherein the adaptive signal processor is a multi-stage median cascaded canceller one or more adaptive weight values.

33. The system according to claim 32 wherein the controller applies the soft weight value by multiplying at least one adaptive weight value of the multi-stage median cascaded canceller by the soft weight value.

34. The system according to claim 26 wherein the adaptive signal processor is a generic cascaded canceller comprising one or more adaptive weight values.

35. The system according to claim 34 wherein the controller applies the soft weight value by multiplying at least one adaptive weight value of the generic cascaded canceller by the soft weight value.

36. The system according to claim 26 wherein the adaptive signal processor is a pseudo-median cascaded canceller comprising one or more adaptive weight values.

37. The system according to claim 36 wherein the controller applies the soft weight value by multiplying at least one adaptive weight value of the pseudo-median cascaded canceller by the soft weight value.

38. The system according to claim 26 wherein the adaptive signal processor is a Multi-Stage Weiner filter comprising one or more adaptive weight values.

39. The system according to claim 38 wherein the Multi-Stage Weiner filter comprises a pre-determined number of stages.

41. The system according to claim 38 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages is selected to be greater than a minimum effective rank of an interference plus noise environment.

42. The system according to claim 38 wherein the Multi-Stage Weiner Filter comprise a plurality of stages such that the number of stages required for optimal convergence is less than the number of degrees of freedom.

43. The system according to claim 38 wherein the controller applies the soft weight value by multiplying at least one Weiner weight value of the Multi-stage Weiner Filter by the soft weight value.

44. The system according to claim 38 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and at least one Gram-Schmidt weight value is applied to the synthesis stage.

45. The system according to claim 44 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and the controller applies the soft weight value by multiplying at least one Gram-Schmidt weight value of the Multi-stage Weiner Filter by the soft weight value.

40. The system according to 38 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages is selected to be equal to an effective rank of an interference plus noise environment.

46. A method for adaptive signal processing, by an adaptive signal processor, the method comprising:

measuring, by one or more sensors of the adaptive signal processor, a signal comprising a plurality of measurements, z and x, in an interference plus noise environment;

inputting the measured signal into an adaptive signal processor wherein a controller retrieves from a memory means a stored number of iterations, such that the operations of the adaptive signal processor are performed a number of times as specified by the number of iterations;

filtering, by the adaptive signal processor, the measured signal to eliminate interference;

applying, by the adaptive signal processor, a soft weight to the algorithm, the soft weight value modifying one or more existing adaptive weight values associated with the adaptive signal processor in accordance with the following equation;

y=z−

(w_sw*w_g)*x,

wherein w_g is adaptive weight value, w_sw is representative of the soft weight complex scalar value; and

outputting, by the adaptive signal processor, the filtered and weighted signal, y.

47. The method of claim 46, wherein the measured signal comprises a target signal, an additive noise, and an interference.

48. The method of claim 46, wherein the filtering comprises a cascaded algorithm.

49. The method of claim 48 wherein the cascaded canceller algorithm is a Gram-Schmidt cascaded canceller algorithm.

50. The method of claim 49 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and applying the soft weight value further comprises multiplying the Gram-Schmidt weight value by the soft weight value.

51. The method of claim 48 wherein the cascaded canceller algorithm is a pseudo-median cascaded canceller algorithm.

52. The method of claim 51 wherein applying the soft weight value further comprises multiplying the one or more adaptive weight values by the soft weight value.

53. The method of claim 48 wherein the cascaded canceller algorithm is a multistage median cascaded canceller.

54. The method of claim 53 wherein applying the soft weight value further comprises multiplying the one or more adaptive weight values by the soft weight value.

55. The method of claim 48 wherein the cascaded canceller algorithm is a generic cascaded canceller algorithm.

56. The method of claim 55 wherein applying the soft weight value further comprises multiplying the one or more adaptive weight values by the soft weight value.

57. The method of claim 46 wherein the filtering comprises a Multi-Stage Weiner Filter algorithm.

58. The method of claim 57 wherein the one or more adaptive weight values is a Weiner weight value, and applying the soft weight value further comprises multiplying the Weiner weight value by the soft weight value.

59. The method of claim 57 wherein the filtering further comprises applying at least one Gram-Schmidt weight in the synthesis stage.

60. The method of claim 59 wherein applying the soft weight value further comprises multiplying the Gram-Schmidt weight by the soft weight value.

61. The method of claim 46 wherein the applying a soft weight value further comprises selecting a soft weight value between 0 and 2, excluding a value of 1.

62. The method of claim 46 wherein applying the soft weight value further comprises selecting a soft weight value such that the pre-determined average value is a signal to power plus noise ratio [SINR] within 3 dB from optimal normalized SINR of 0 dB, the soft weight value selected in less than 2N training samples, where N is the number of degrees of freedom.

63. An adaptive signal processor comprising:

a filter;

a controller for retrieving from a memory means a stored number of iterations, such that operations of the controller, the operations of the adaptive signal processor, and the operations of a soft weight generator are performed a number of times as specified by the number of iterations;

a soft weight generator for modifying one or more adaptive weight values associated with the adaptive signal processor; and

an output y, determined in accordance with the following equation;

y=z−

(w_sw*w_g)*x,

wherein x and z are input signals, including the measured signal, into the filter, w_g is conjugated adaptive weight value, w_sw is representative of the soft weight complex scalar value.

64. The adaptive signal processor of claim 63 wherein the filter further comprises an input from the soft weight generator.

65. The adaptive signal processor of claim 64 wherein the soft weight generator generates a soft weight value between 0 and 2, excluding a value of 1.

66. The adaptive signal processor of claim 63, wherein the filter further comprises a cascaded canceller that applies the soft weight value.

67. The adaptive signal processor of claim 66 wherein the cascaded canceller further comprises a Gram-Schmidt cascaded canceller.

68. The adaptive signal processor of claim 67 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and applying the soft weight value further comprises multiplying the Gram-Schmidt weight value by the soft weight value.

69. The adaptive signal processor of claim 66 wherein the cascaded canceller further comprises a pseudo-median cascaded canceller.

70. The adaptive signal processor of claim 69 wherein applying the soft weight value further comprises multiplying the one or more adaptive weight values by the soft weight value.

71. The adaptive signal processor of claim 66 wherein the cascaded canceller further comprises a median multi-stage cascaded canceller.

72. The adaptive signal processor of claim 71 wherein applying the soft weight value further comprises multiplying the one or more adaptive weight values by the soft weight value.

73. The adaptive signal processor of claim 66 wherein the cascaded canceller further comprises a generic cascaded canceller.

74. The adaptive signal processor of claim 73 wherein applying the soft weight value further comprises multiplying the one or more adaptive weight values by the soft weight value.

75. The adaptive signal processor according to claim 65 wherein the filter comprises a Multi-stage Weiner filter that applies the soft weight value.

76. The adaptive signal processor of claim 75 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages is selected to be equal to an effective rank of an interference plus noise environment.

77. The adaptive signal processor of claim 75 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages is selected to be greater than a minimum effective rank of an interference plus noise environment.

78. The adaptive signal processor of claim 75 wherein the Multi-Stage Weiner Filter comprises a plurality of stages such that the number of stages required for optimal convergence is less than the number of degrees of freedom.

79. The adaptive signal processor of claim 75 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and at least one Gram-Schmidt weight value is applied in the synthesis stage.

80. The adaptive signal processor of claim 79 wherein the one or more adaptive weight values is a Gram-Schmidt weight value, and applying the soft weight value further comprises multiplying the Gram-Schmidt weight value by the soft weight value.

81. The adaptive signal processor of claim 75 wherein the one or more adaptive weight values is a Weiner weight value, and applying the soft weight value further comprises multiplying the Weiner weight value by the soft weight value.

82. The adaptive signal processor of claim 65 wherein the soft weight value is selected such that the pre-determined average value is a signal to power plus noise ratio [SINR] within 3 dB from optimal normalized SINR of 0 dB, the soft weight value selected in less than 2N training samples, where N is the number of degrees of freedom.