Data processing for automatic modulation classification of wireless signals
Systems and methods for classifying a modulation scheme of a wireless signal are described. In some embodiments, a system receives a wireless signal modulated based on a modulation scheme having a constant modulus. The system can generate a resampled signal from the wireless signal based on features extracted from the wireless signal and perform blind equalization on the resampled signal based on a constant modulus criterion to generate an equalized signal. Then, the system can cause a modulation classifier to classify the received wireless signal to a modulation scheme from a plurality of predetermined modulation schemes based on the equalized signal. By preconditioning the wireless signal to reduce feature variability imparted by a propagation channel onto the wireless signal, the system can increase the classification accuracy of the modulation classifier.
1. A method for classifying a modulation scheme of a wireless signal, comprising:
receiving a modulated wireless signal;
estimating a center frequency of the wireless signal;
estimating a bandwidth of the wireless signal;
generating a resampled signal from the received modulated wireless signal wherein a rate of the resampled signal is based on the estimated center frequency and the estimated bandwidth of the wireless signal, wherein generating a resampled signal from the wireless signal comprises performing blind equalization in an iterative manner, and wherein during each iteration, the resampled signal is convolved with a plurality of equalization taps and a convolved signal envelope is used to determine whether to adjust a plurality of equalization weights; and
classifying the received wireless signal using a modulation classifier to a modulation scheme based on the resampled signal.
2. The method of claim 1 , wherein estimating the center frequency of the wireless signal comprises generating a power spectrum and estimating the center frequency based on the generated power spectrum.
3. The method of claim 2 , wherein estimating the center frequency comprises applying a Costas loop estimation process.
4. The method of claim 2 , wherein estimating the center frequency comprises applying an ambiguity function estimation process.
5. The method of claim 2 , wherein estimating the center frequency comprises applying an autoregressive model estimation process.
6. The method of claim 2 , wherein estimating the center frequency comprises applying a subspace tracking estimation process.
7. The method of claim 1 , wherein determining the estimated bandwidth of the wireless signal comprises generating a power spectrum and estimating the bandwidth based on the generated power spectrum.
8. The method of claim 7 , wherein estimating the bandwidth comprises applying a Fourier transform estimation process.
9. The method of claim 7 , wherein estimating the bandwidth comprises applying a wavelet transform estimation process.
10. The method of claim 7 , wherein estimating the bandwidth comprises applying a candidate signal matched filter bank estimation process.
11. The method of claim 1 , wherein the method comprises determining a symbol rate and an oversampling rate of the wireless signal, and wherein the oversampling rate is determined based on the symbol rate.
12. The method of claim 11 , wherein determining the symbol rate comprises performing a cyclic power spectrum analysis.
13. The method of claim 11 , wherein determining the symbol rate comprises applying a cyclo-stationary technique different than a cyclic power spectrum analysis.
14. The method of claim 11 , wherein determining the symbol rate comprises applying a wavelet transform symbol rate estimation process.
15. The method of claim 1 , wherein blind equalization is performed in an iterative manner to estimate a plurality of equalization weights for a plurality of corresponding equalization taps.
16. The method of claim 1 , wherein determining whether to adjust the plurality of equalization weights comprises:
comparing a magnitude of the convolved signal with a predetermined value, and
determining that blind equalization is complete if the difference between the magnitude and the predetermined value is less than a threshold value.
17. The method of claim 1 , wherein determining whether to adjust the plurality of equalization weights comprises:
determining whether the difference between a magnitude of the convolved signal and a predetermined value is converging or diverging from a previous iteration, and
if the error value is diverging, regenerating a resampled signal based on one or more of a re-estimated center frequency and a re-estimated signal bandwidth of the wireless signal.
18. The method of claim 1 , wherein adjusting the plurality of equalization weights comprises applying a least means squared (LMS) algorithm filter to recalculate the plurality of equalization rates.
19. The method of claim 1 , wherein adjusting the plurality of equalization weights comprises applying a recursive least squares (RLS) algorithm filter to recalculate the plurality of equalization rates.
20. The method of claim 1 , wherein adjusting the plurality of equalization weights comprises applying a Kalman filter to recalculate the plurality of equalization rates.
21. The method of claim 1 , wherein adjusting the plurality of equalization weights comprises applying an auto-regressive moving-average model filter to recalculate the plurality of equalization rates.
22. The method of claim 1 , wherein the wireless signal comprises an underwater acoustic signal.
23. The method of claim 1 , wherein the modulation classifier comprises a supervised classifier.
24. The method of claim 23 , wherein the supervised classifier comprises a decision trees classifier, a k-nearest neighbors (k-NN) classifier, a neural networks classifier, a linear classifier, or a margin classifier.
25. The method of claim 1 , wherein the modulation scheme comprises one or more modulation types including frequency modulation (FM), phase modulation (PM), frequency-shift keying (FSK), phase-shift keying (PSK), or minimum-shift keying (MSK).
26. A system for classifying a modulation scheme of a wireless signal, comprising:
a receiver configured to receive a wireless signal;
a signal processor configured to:
estimate a center frequency of the wireless signal;
estimate a bandwidth of the wireless signal; and
generate a resampled signal from the received wireless signal wherein a rate of the resampled signal is based on the estimated center frequency and the estimated bandwidth of the wireless signal, wherein generating a resampled signal from the wireless signal comprises performing blind equalization in an iterative manner, and wherein during each iteration, the resampled signal is convolved with a plurality of equalization taps and a convolved signal envelope is used to determine whether to adjust a plurality of equalization weights; and
a modification classifier configured to classify the received wireless signal to a modulation scheme based on the resampled signal.