Method and apparatus for compressed sensing

• US 7,289,049 B1
• Filed: 08/21/2006
• Issued: 10/30/2007
• Est. Priority Date: 08/21/2006
• Status: Expired due to Fees

• Current Assignee: L-3 Communications Corporation (L3Harris Technologies, Inc.)
• Original Assignee: L-3 Communications Integrated Systems LP (L3Harris Technologies, Inc.)
• Inventors: Fudge, Gerald Lothair, Wood, Mark L., Yeh, Chen-Chu Alex
• Primary Examiner(s): Williams; Howard L.
• Application Number:
  US11/465,826
• Time in Patent Office: 435 Days
• Field of Search: 341/122, 341/123, 341/155, 702 64- 79
• US Class Current: 341/123
• CPC Class Codes: H03M 1/1245 Details of sampling arrangements or methods

Abstract

Embodiments of the present invention provide a method and apparatus for compressed sensing. The method generally comprises forming a first compressed sensing matrix utilizing a first set of time indices corresponding to a first sampling rate, forming a second compressed sensing matrix utilizing a plurality of frequencies and a second set of time indices corresponding to a second sampling rate, forming a combined compressed sensing matrix from the first compressed sensing matrix and the second compressed sensing matrix, and reconstructing at least a portion of the input signal utilizing the combined compressed sensing matrix. The first and second sampling rates are each less than the Nyquist sampling rate for the input signal.

Claims

1. A method of reconstructing an input signal, the method comprising the steps of:

(a) forming a first compressed sensing matrix utilizing a first set of time indices corresponding to a first sampling rate, the first sampling rate being less than the Nyquist sampling rate for the input signal;

(b) forming a second compressed sensing matrix utilizing a plurality of frequencies and a second set of time indices corresponding to a second sampling rate, the second sampling rate being less than the Nyquist sampling rate for the input signal;

(c) forming a combined compressed sensing matrix from the first compressed sensing matrix and the second compressed sensing matrix; and

(d) reconstructing at least a portion of the input signal utilizing the combined compressed sensing matrix.

2. The method of claim 1, further comprising the steps of:

(e) sampling the input signal at the first sampling rate to form a first measurement vector,

(f) sampling the input signal at the frequencies and the second sampling rate to form a second measurement vector, and

(g) forming a combined measurement vector utilizing the first measurement vector and the second measurement vector.

3. The method of claim 2, wherein step (d) includes reconstructing at least a portion of the input signal utilizing the combined compressed sensing matrix and the combined measurement vector.

4. The method of claim 1, wherein step (a) includes selecting sampling times for inclusion in the first set of time indices.

5. The method of claim 1, wherein the first sampling rate and the second sampling rate are substantially equal.

6. The method of claim 1, wherein the sum of the first sampling rate and the second sampling rate is less than the Nyquist sampling rate for the input signal.

7. The method of claim 1, wherein step (b) includes forming the second compressed sensing matrix utilizing the second set of time indices and a Fourier transform matrix generated utilizing a discrete time Fourier transform of an identity matrix.

8. The method of claim 1, wherein the plurality of frequencies correspond to a plurality of narrowband filters.

9. The method of claim 1, wherein step (d) includes reconstructing the input signal utilizing the combined compressed sensing matrix to form a full-bandwidth representation of at least a portion of the input signal at the Nyquist sampling rate.

10. An apparatus for reconstructing an input signal, the apparatus comprising:

a memory operable to store data corresponding to a plurality of frequencies, a first set of time indices corresponding to a first sampling rate, a second set of time indices corresponding to a second sampling rate, the first and second sampling rates each being less than the Nyquist sampling rate for the input signal; and

a processor coupled with the memory to receive data therefrom, the processor operable to—

form a first compressed sensing matrix utilizing data corresponding to the first set of time indices,

form a second compressed sensing matrix utilizing data corresponding to the plurality of frequencies and the second set of time indices,

form a combined compressed sensing matrix from the first compressed sensing matrix and the second compressed sensing matrix, and

reconstruct at least a portion of the input signal utilizing the combined compressed sensing matrix.

11. The apparatus of claim 10, wherein the memory is further operable to store data corresponding to the input signal sampled at the plurality of frequencies, the first sampling rate, and the second sampling rate.

12. The apparatus of claim 11, wherein the processor is further operable to—

generate a first measurement vector utilizing data corresponding to the input signal sampled at the first sampling rate,

generate a second measurement vector utilizing data corresponding to the input signal sampled at the second sampling rate,

form a combined measurement vector utilizing the first measurement vector and the second measurement vector, and

reconstruct the input signal utilizing the combined compressed sensing matrix and the combined measurement vector.

13. The apparatus of claim 10, wherein the sum of the first sampling rate and the second sampling rate is less than the Nyquist sampling rate for the input signal.

14. The apparatus of claim 10, wherein the processor reconstructs the input signal utilizing the combined compressed sensing matrix to form a full-bandwidth representation of at least a portion of the input signal at the Nyquist sampling rate.

15. The apparatus of claim 10, wherein the memory is integral with the processor.

16. An apparatus for reconstructing an input signal, the apparatus comprising:

a receiving element operable to receive the input signal;

a wideband filter coupled with the receiving element to filter the input signal to a first range of frequencies and provide a wideband output signal corresponding to the filtered input signal;

a plurality of narrowband filters coupled with the receiving element, each narrowband filter being operable to filter the input signal to a second range of frequencies, each second range of frequencies being less than the first range of frequencies, the plurality of narrowband filters operable to provide a narrowband output signal corresponding to the sum of the outputs of the narrowband filters;

a sampling element coupled with the wideband filter and the narrowband filters to receive the wideband output signal and the narrowband output signal, the sampling element operable to sample the wideband output signal at a first sampling rate and sample the narrowband output signal at a second sampling rate to provide at least one sampled signal, the first and second sampling rates each being less than the Nyquist sampling rate for the input signal; and

a processor coupled with the sampling element and operable to—

form a first compressed sensing matrix utilizing a first set of time indices corresponding to the first sampling rate,

form a second compressed sensing matrix utilizing a plurality of frequencies corresponding to the narrowband filters and a second set of time indices corresponding to the second sampling rate,

form a combined compressed sensing matrix from the first compressed sensing matrix and the second compressed sensing matrix, and

reconstruct at least a portion of the input signal utilizing the combined compressed sensing matrix.

17. The apparatus of claim 16, wherein first sampling rate and the second sampling rate are equal.

18. The apparatus of claim 16, wherein the sampling element comprises a first sampling element coupled with the wideband filter to receive and sample the wideband output signal at the first sampling rate and a second sampling element coupled with the narrowband filter to receive and sample the narrowband filter output signal at the second sampling rate.

19. The apparatus of claim 16, wherein the sum of the first sampling rate and the second sampling rate is less than the Nyquist sampling rate for the input signal.

20. The apparatus of claim 16, wherein the processor is further operable to—

form a measurement vector utilizing the sampled signal, and

reconstruct at least a portion of the input signal utilizing the combined compressed sensing matrix and the measurement vector.

21. The apparatus of claim 16, wherein the sampling element is operable to provide a sampled narrowband output signal and a sampled wideband output signal and the processor is further operable to—

form a first measurement vector utilizing the sampled wideband output signal,

form a second measurement vector utilizing the sampled narrowband output signal,

form a combined measurement vector utilizing the first and second measurement vectors, and

reconstruct the input signal utilizing the combined compressed sensing matrix and the combined measurement vector.

22. The apparatus of claim 16, wherein the processor is operable to form the second compressed sensing matrix utilizing the plurality of frequencies, the second set of time indices, and a Fourier transform matrix generated utilizing a discrete time Fourier transform of an identity matrix.

23. The apparatus of claim 16, wherein the processor reconstructs at least a portion of the input signal utilizing the combined compressed sensing matrix to form a full-bandwidth representation of the input signal at the Nyquist sampling rate.

24. The apparatus of claim 16, wherein the processor is further operable to select at least one of the narrowband filters to determine which narrowband filters are utilized to provide the narrowband output signal and form the second compressed sensing matrix.

25. An apparatus for sampling an input signal, comprising:

a receiving element operable to receive the input signal;

a wideband filter coupled with the receiving element to filter the input signal to a first range of frequencies and provide a wideband output signal corresponding to the filtered input signal;

a plurality of narrowband filters coupled with the receiving element, each narrowband filter being operable to filter the input signal to a second range of frequencies, each second range of frequencies being less than the first range of frequencies, the plurality of narrowband filters operable to provide a narrowband output signal corresponding to the sum of the outputs of the narrowband filters;

a sampling element coupled with the wideband filter and the narrowband filters to receive the wideband output signal and the narrowband output signal, the sampling element operable to sample the wideband output signal at a first sampling rate and sample the narrowband output signal at a second sampling rate to provide at least one sampled signal; and

a processor coupled with the sampling element and operable to process the sampled signal.

26. The apparatus of claim 25, wherein the first and second sampling rates are each less than the Nyquist sampling rate for the input signal and the processor is further operable to—

form a first compressed sensing matrix utilizing a first set of time indices corresponding to the first sampling rate,

form a second compressed sensing matrix utilizing a plurality of frequencies corresponding to the narrowband filters and a second set of time indices corresponding to the second sampling rate,

form a combined compressed sensing matrix from the first compressed sensing matrix and the second compressed sensing matrix, and

reconstruct at least a portion of the input signal utilizing the combined compressed sensing matrix.

27. The apparatus of claim 26, wherein the first and second sampling rates are equal.

28. The apparatus of claim 26, wherein the sum of the first sampling rate and the second sampling rate is less than the Nyquist sampling rate for the input signal.

29. The apparatus of claim 25, wherein the processor is operable to perform direct detection on the sampled signal.

30. A computer program for reconstructing an input signal, the computer program stored on a computer-readable medium for operating a processor and comprising:

a code segment operable to form a first compressed sensing matrix utilizing a first set of time indices corresponding to a first sampling rate, the first sampling rate being less than the Nyquist sampling rate for the input signal;

a code segment operable to form a second compressed sensing matrix utilizing a plurality of frequencies and a second set of time indices corresponding to a second sampling rate, the second sampling rate being less than the Nyquist sampling rate for the input signal;

a code segment operable to form a combined compressed sensing matrix from the first compressed sensing matrix and the second compressed sensing matrix; and

a code segment operable to reconstruct at least a portion of the input signal utilizing the combined compressed sensing matrix.

31. The computer program of claim 30, further comprising—

a code segment operable to acquire data corresponding to sampling of the input signal at the first sampling rate and form a first measurement vector utilizing the data acquired at the first sampling rate,

a code segment operable to acquire data corresponding to sampling of the input signal at the second sampling rate and form a second measurement vector utilizing the data acquired at the second sampling rate, and

a code segment operable to form a combined measurement vector utilizing the first measurement vector and the second measurement vector.

32. The computer program of claim 31, wherein at least a portion of the input signal is reconstructed utilizing the combined compressed sensing matrix and the combined measurement vector.

33. The computer program of claim 30, wherein the sum of the first sampling rate and the second sampling rate is less than the Nyquist sampling rate for the input signal.

34. The computer program of claim 30, wherein the first sampling rate and the second sampling rate are equal.

35. The computer program of claim 30, wherein the second compressed sensing matrix is formed utilizing the plurality of frequencies, the second set of time indices, and a Fourier transform matrix generated utilizing a discrete time Fourier transform of an identity matrix.

36. The computer program of claim 30, wherein at least a portion of the input signal is reconstructed utilizing the combined compressed sensing matrix to form a full-bandwidth representation of the signal at the Nyquist sampling rate.

37. The computer program of claim 30, wherein the plurality of frequencies correspond to a plurality of narrowband filters.