EFFICIENT DETECTION ALGORITHM SYSTEM FOR A BROAD CLASS OF SIGNALS USING HIGHER-ORDER STATISTICS IN TIME AS WELL AS FREQUENCY DOMAINS

• US 20110131260A1
• Filed: 12/22/2009
• Published: 06/02/2011
• Est. Priority Date: 06/16/2006
• Status: Active Grant

• Current Assignee: BAE Systems Information and Electronic Systems Integration Incorporated (BAE Systems Plc)
• Original Assignee: BAE Systems Information and Electronic Systems Integration Incorporated (BAE Systems Plc)
• Inventors: Mody, Apurva N.
• Granted Patent: US 8,688,759 B2
• Time in Patent Office: Days
• US Class Current: 708/5
• CPC Class Codes: H04L 27/0006 Assessment of spectral gaps suitable for allocating digitally modulated signals, e.g. for carrier allocation in cognitive radio for spectrum sharing between different networks H04W16/14

Abstract

An algorithm system to detect a broad class of signals in Gaussian noise using higher-order statistics. The algorithm system detects a number of different signal types. The signals may be in the base-band or the pass-band, single-carrier or multi-carrier, frequency hopping or non-hopping, broad-pulse or narrow-pulse etc. In a typical setting this algorithm system provides an error rate of 3/100 at a signal to noise ratio of 0 dB. This algorithm system gives the time frequency detection ratio that may be used to determine if the detected signal falls in Class Single-Carrier of Class Multi-Carrier. Additionally this algorithm system may be used for a number of different applications such as multiple signal identification, finding the basis functions of the received signal and the like.

Claims

1. A method for implementation of a Spectrum Sensing Function (SSF) for detecting signals in Gaussian noise, wherein Higher Order Statistics (HOS) are applied to segments of received signals in at least one of time and frequency domains comprising the steps of:

moving to a particular portion of a frequency spectrum;

applying a band pass filter;

applying a low noise amplifier to output of said band pass filter;

adjusting gain of said amplified output of said band pass filter;

collecting waveforms in said portion of a frequency spectrum;

downconverting said collected waveforms;

applying an analog to digital conversion;

applying a low pass filter;

converting to focus on a spectrum of interest;

sampling to adjust a sampling rate;

applying serial to parallel conversion to convert a stream of samples;

applying a Fast Fourier Transform (FFT);

detecting at least one signal using said Higher Order Statistics;

classifying a segment as belonging to Class Signal or Class Noise; and

identifying said at least one signal.

2. The method of claim 1, comprising the step of:

processing said at least one signal in said time domain to detect at least one signal.

3. The method of claim 1, comprising the step of:

processing said at least one signal in said frequency domain to detect at least one signal.

4. The method of claim 1, comprising the step of:

processing said at least one signal in said time as well as said frequency domains and combining results to detect at least one signal.

5. The method of claim 1 comprising the step of:

dividing received data sample stream into smaller segments, whereby said Higher Order Statistics (HOS) signal detection is carried out for each of said segments.

6. The method of claim 1, comprising:

dividing said data segments into real and imaginary parts, wherein R is the number of moments (m<sub>r—

real</sub>, m<sub>r—

imaginary</sub>) and cumulants (c<sub>r—

real</sub>, c<sub>r—

imaginary</sub>) of order greater than two available for computation for said real and said imaginary parts of each of said data segments respectively;

choosing a value for probability step parameter (δ

) between zero and one;

setting Psignal_real and Psignal_imaginary to 0.5;

choosing a value for fine threshold parameter (γ

) greater than zero, wherein said fine threshold parameter γ

is used to control probability of false alarm P_FA and probability of detection P_D;

computing all R+2 moments and cumulants, wherein for r=3 to (R+2), if |c<sub>r—

real</sub>| is less than γ

|m<sub>2—

real</sub>|^r/2, then P<sub>Signal—

real</sub> equals P<sub>Signal—

real</sub>−

δ

, if |c<sub>r—

real</sub>| is greater than or equal to γ

|m<sub>2—

real</sub>|^r/2, then P<sub>Signal—

real</sub> equals P<sub>Signal—

real</sub>+δ

, and wherein if |c<sub>r—

imaginary</sub>| is less than γ

|m<sub>2—

imaginary</sub>|^r/2, then P<sub>Signal—

imaginary</sub> equals P<sub>Signal—

imaginary</sub>−

δ

, if |c<sub>r—

imaginary</sub>| is greater than or equal to γ

|m<sub>2—

imaginary</sub>|^r/2, then P<sub>Signal—

imaginary</sub> equals P<sub>Signal—

imaginary</sub>+δ

, and wherein P_Signal equals aP<sub>Signal—

real</sub> plus bP<sub>Signal—

imaginary</sub>, wherein a and b are weight parameter coefficients;

assigning said data sample segment to Class Signal if P_Signal is greater than or equal to 0.5; and

assigning said data sample segment to Class Noise if P_Signal is less than 0.5 and no signal is detected.

7. The method of claim 6, wherein a Fourier Transform is applied to said data segments to convert said data segments into said frequency domain.

8. The method of claim 7, wherein said Fourier Transform is a Fast Fourier Transform (FFT).

9. The method of claim 1, wherein results of signal detection in said time and said frequency domains are combined to determine if said data segment belongs to Class Signal or Class Noise;

and

wherein combining is defined by Psignal=c Psignal_time+d Psignal_frequency, where c and d are coefficients such that c+d=1.

10. The method of claim 1 further comprising:

pre-processing, said preprocessing comprising at least one of filtering, noise whitening, down-conversion, up-conversion, frequency shift, frequency translation, re-sampling, down-sampling, up-sampling, signal conditioning, wherein said preprocessing is applied to said data segments before computing said HOS of said data segments in at least one of said time and said frequency domains, and wherein sequence of said preprocessing is alterable.

11. The method of claim 1, wherein said received signals are in base-band.

12. The method of claim 1, wherein said received signals are in pass-band.

13. The method of claim 1, wherein said received signals are single-carrier.

14. The method of claim 1, wherein said received signals are multi-carrier.

15. The method of claim 1, wherein said received signals are frequency hopping.

16. The method of claim 1, wherein said received signals are non-frequency hopping.

17. The method of claim 1, wherein said received signals are at least one of Direct Sequence Spread Spectrum and Multi-Carrier Spread Spectrum.

18. The method of claim 1, wherein said received signals are non-Gaussian in said time domain.

19. The method of claim 1, wherein said received signals are non-Gaussian in said frequency domain.

20. The method of claim 1, wherein detecting said at least one signal using said higher order statistics further comprises:

adjusting a fine threshold parameter (γ

); and

reclassifying said at least one signal.

21. The method of claim 1, comprising:

choosing all cumulants greater than two for computation of signal detection probabilities.

22. The method of claim 1, wherein one SSF interfaces with at least one of a Spectrum Manager or a Cognitive Engine.

23. The method of claim 1, wherein a plurality of SSFs interface with at least one of a Spectrum Manager or a Cognitive Engine.

24. The method of claim 1, wherein a plurality of SSFs interface with a plurality of Spectrum Managers or Cognitive Engines.

25. The method of claim 1, wherein the decision on whether a signal is detected in said channel or sub-band is used by a Cognitive Engine or a Spectrum Manager to further arrive at a decision on whether said channel or sub-band may be used for communications.

26. The method of claim 1, wherein said SSF reports total signal energy in every channel at least one of a sub-band, a Power Spectral Density (PSD), an Average Channel Power (ACP), a Received Signal Strength Indicator (RSSI), a Carrier to Interference plus Noise Ratio (CINR), or a Field Strength so that at least one of a Spectrum Manager or a Cognitive Engine can make a better decision on prioritizing use of said channel or sub-band.

27. The method of claim 1, comprising:

choosing a probability step parameter (δ

) equal to one-half the inverse of a number of moments and cumulants of order greater than two available for computation of real and imaginary parts of each segment of said received signal.

28. The method of claim 27, comprising:

choosing a subset of said cumulants for computation of signal detection probabilities.

35. The method of claim 6, wherein Class Signal or Class Noise determination is performed in said frequency domain by computing Psignal_frequency, where Psignal_frequency=Psignal computed in said frequency domain.

36. The method of claim 1, comprising:

choosing a subset of cumulants greater than two for computation of signal detection probabilities.

37. The method of claim 1, comprising:

choosing even ordered cumulants for computation of signal detection probabilities.

38. The method of claim 1, comprising:

choosing a subset of even ordered cumulants for computation of signal detection probabilities.

39. The method of claim 26, wherein said channel or said sub-band with lowest PSD, RSSI, or Field Strength is selected for communications.

40. The method of claim 1, wherein classifying said detected signal into Class Single Carrier or Class Multi Carrier is based on a Time Frequency Detection Ratio.

29. A method for implementation of a spectrum sensing function for detecting signals in Gaussian noise, wherein Higher Order Statistics (HOS) are applied to segments of received signals in at least one of time and frequency domains comprising the steps of:

moving to a particular portion of a frequency spectrum;

dividing received data sample stream into smaller segments, whereby said HOS signal detection is carried out for each of said segments;

pre-processing, said preprocessing comprising at least one of filtering, noise whitening, down-conversion, up-conversion, frequency shift, frequency translation, re-sampling, down-sampling, up-sampling, signal conditioning, wherein said preprocessing is applied to said data segments before computing said HOS of said data segments in at least one of said time and said frequency domains, and wherein sequence of said preprocessing is alterable;

applying a band pass filter;

applying a low noise amplifier to output of said band pass filter;

adjusting gain of said amplified output of said band pass filter;

collecting waveforms in said portion of a frequency spectrum;

downconverting said collected waveforms;

applying an analog to digital conversion;

applying a low pass filter;

converting to focus on a spectrum of interest;

sampling to adjust a sampling rate;

applying serial to parallel conversion to convert a stream of samples;

applying a Fast Fourier Transform (FFT), wherein said FFT is applied to said data segments to convert said data segments into said frequency domain;

detecting at least one signal using said Higher Order Statistics, wherein said HOS signal detection is in said time and said frequency domains;

dividing said data segments into real and imaginary parts, wherein R is the number of moments (m<sub>r—

real</sub>, m<sub>r—

imaginary</sub>) and cumulants (c<sub>r—

real</sub>, c<sub>r—

imaginary</sub>) of order greater than two available for computation for said real and said imaginary parts of each of said data segments respectively;

choosing a value for probability step parameter (δ

) equal to one-half the inverse of a number of moments and cumulants of order greater than two available for computation of real and imaginary parts of each segment of said received signal;

choosing a value for fine threshold parameter (γ

) greater than zero, wherein said fine threshold parameter γ

is used to control probability of false alarm P_FA and probability of detection P_D;

setting Psignal_real and Psignal_imaginary to 0.5;

computing all R+2 moments and cumulants, wherein for R=3 to (R+2), if |c<sub>r—

real</sub>| is less than γ

|m<sub>2—

real</sub>|^r/2, then P<sub>Signal—

real</sub> equals P<sub>Signal—

real</sub>−

δ

, if c<sub>r—

real</sub> is greater than or equal to γ

|m<sub>2—

real</sub>|^r/2, then P<sub>Signal—

real</sub> equals P<sub>Signal—

real</sub>+δ

, and wherein if |c<sub>r—

imaginary</sub>| is less than γ

|m<sub>2—

imaginary</sub>|_r/2, then P<sub>Signal—

imaginary</sub> equals P<sub>Signal—

imaginary</sub>−

δ

, if |c<sub>r—

imaginary</sub>| is greater than or equal to γ

|m<sub>2—

imaginary</sub>|^r/2, then P<sub>Signal—

imaginary</sub> equals P<sub>Signal—

imaginary</sub>+δ

, and wherein P_Signal equals aP<sub>Signal—

real</sub> plus bP<sub>Signal—

imaginary</sub>, wherein a and b are weight parameter coefficients;

assigning said data sample segment to Class Signal if P_Signal is greater than or equal to 0.5;

assigning said data sample segment to Class Noise if P_Signal is less than 0.5 and no signal is detected;

adjusting a fine threshold parameter (γ

);

reclassifying said at least one signal;

classifying a segment as belonging to Class Signal or Class Noise, wherein results of said HOS signal detection in said time and said frequency domains are combined to determine if said data segment belongs to Class Signal or Class Noise; and

identifying said at least one signal.

30. A system for Spectrum Sensing and signal identification wherein Higher Order Statistics (HOS) are applied to segments of received signals in time and frequency domains comprising:

signal detection and identification classification modules configured to perform the steps of;

moving to a particular portion of a frequency spectrum;

applying a band pass filter;

applying a low noise amplifier to output of said band pass filter;

adjusting gain of said amplified output of said band pass filter;

collecting waveforms present in said spectrum;

downconverting said collected waveforms;

applying an analog to digital conversion in an analog to digital converter;

first filtering down-converted signal through an image rejection first Low Pass (LP) filter, wherein an image of said downconverted signal is suppressed;

upconverting said first filtered signal, wherein a video carrier would be shifted closer to 0 Hertz frequency;

second filtering said upconverted signal;

downsampling said second filtered signal;

converting samples of said downsampled signal from serial to parallel in a serial to parallel converter;

collecting said samples;

storing said samples in a buffer;

applying a Fast Fourier Transform (FFT);

determining higher order moments and cumulants of real and imaginary portions of said stored samples;

calculating signal probability; and

classifying said received signal.

31. A method for classifying a Denial of Service (DoS) signal comprising the steps of:

determining bit error rate degradation of a received signal;

determining the Carrier to Interference plus Noise Ratio (CINR);

determining the Received Signal Strength Indication (RSSI);

performing signal or noise detection on said received signal using higher order statistics (HOS);

detecting time and frequency domain components of said received signal; and

identifying Gaussianity whereby said DoS signal is classified from results of said detecting step.

32. A method for signal identification comprising the steps of:

moving to a particular portion of a frequency spectrum;

applying a band pass filter;

collecting waveforms present in said spectrum;

downconverting said collected waveforms;

applying an analog to digital conversion;

first filtering down-converted signal through an image rejection first Low Pass (LP) filter, wherein an image of said downconverted signal is suppressed;

upconverting said first filtered signal, wherein a characteristic frequency component of said signal would be shifted closer to 0 Hertz frequency;

second filtering said upconverted signal;

downsampling said second filtered signal;

converting samples of said downsampled signal from serial to parallel;

collecting said samples;

storing said samples in a buffer;

applying a Fast Fourier Transform (FFT);

determining higher order moments and cumulants of real and imaginary portions of said stored samples;

calculating signal probability;

classifying received signal; and

choosing a probability step parameter (δ

) equal to one-half the inverse of a number of moments and cumulants of order greater than two available for computation of real and imaginary parts of each segment of said received signal.

33. The method of claim 32 comprising:

choosing all cumulants greater than two for computation of signal detection probabilities.

34. The method of claim 32 comprising:

choosing a subset of cumulants for computation of signal detection probabilities.