Apparatus and method for sampling and reconstruction of wide bandwidth signals below Nyquist rate

• US 8,861,588 B2
• Filed: 03/09/2012
• Issued: 10/14/2014
• Est. Priority Date: 04/04/2011
• Status: Expired due to Fees

• Current Assignee: The United States of America As Represented By The Secretary of Agriculture
• Original Assignee: The United States Of America As Represented By The Secretary Of The Army
• Inventors: Nguyen, Lam Huy, Tran, Trac D
• Primary Examiner(s): Ahn, Sam K
• Assistant Examiner(s): Murillo Garcia, Fabricio R
• Application Number:
  US13/416,062
• Publication Number:
  US 20120250748A1
• Time in Patent Office: 949 Days
• Field of Search: 375/240, 375/224, 375/316, 341/155
• US Class Current: 375/240
• CPC Class Codes: G01S 13/0209 Systems with very large relative bandwidth, i.e. larger than 10 %, e.g. baseband, pulse, carrier-free, ultrawideband
• CPC Class Codes: G01S 13/9017 with time domain processing of the SAR signals in azimuth G01S13/9005 takes precedence

Abstract

A method and system of reconstructing data signals from one of incomplete measurements comprising a receiver for receiving data signals, an ADC system operatively connected to the receiver that digitizes the received data signal at a slower rate than the Nyquist rate to obtain sparse measurements; first and second dictionaries comprising a plurality of time shifted responses recovered from the data signal; the first dictionary comprising time shifted versions of the previously observed data signals which are sampled at or above the Nyquist minimum sample rate; the second dictionary comprising time shifted versions are sampled below the Nyquist minimum, and at least one processor for reconstruction of the waveform signals by transforming the sub-Nyquist digitized output using the first and second dictionaries to produce the data signal.

Claims

1. A system of reconstructing data signals from one of incomplete, compressed or sparse measurements comprising:

a receiver for receiving data signals from a target area,

an analog to digital conversion system operatively connected to the receiver, the analog to digital conversion system digitizes a received data signal at a slower rate than a Nyquist minimum sampling rate to obtain a sub-Nyquist digitized output;

first and second dictionaries for storage of a plurality of time shifted data recovered from previously transmitted signals;

the first dictionary operating to store time shifted versions of previously transmitted signals which are sampled above the Nyquist minimum sample rate;

the second dictionary operating to store time shifted versions of previously transmitted signals which are sampled below the Nyquist minimum sample rate, and

at least one processor for reconstruction of waveform signals;

the at least one processor operating to transform the sub-Nyquist digitized output from the analog to digital conversion system into an approximation of the received data signal using the first and second dictionaries to produce reconstruct the data signal.

2. The system of claim 1 wherein the receiver comprises a single sensor that provides a received analog signal to the analog to digital conversion system and regular uniform sampling of the analog to digital conversion system records undersampled signal measurements and, wherein the at least one processor operates to recover the received signal using one of orthogonal matching pursuit, basis pursuit, gradient projection, matching pursuit, orthogonal matching pursuit, regularized matching pursuit or subspace pursuit.

3. The system of claim 2 wherein the first dictionary comprises transmitted signals in which a signal to be recovered x can be represented as x=Sα

where x represents column vectors containing Nyquist length records and matrix S is a redundant matrix whose columns contain time shifted versions of the transmitted signal and α

is the sparse vector of weighting coefficients wherein significant coefficients reveal potential phase-shift and amplitude information of significant targets, and wherein the second dictionary is a sub-Nyquist dictionary comprising transmitted signals which can be represented as y=S′

α

, where S′

=Φ

S is the sub-Nyquist version of S, and Φ

is a sampling matrix and y=Φ

x.

4. The system of claim 3 wherein the second dictionary can be represented by S′

and is a corresponding sub-Nyquist version of the first dictionary represented by S constructed such that

S′

_i=Φ

S_i

where S_i is the i^th column of S while S′

_i is the i^th column of S′

, and S′

is a (K×

N) matrix that comprises time-shifted versions of a transmitted waveform similar to the formation of S except that each of the time-shifted column S′

_i in S′

is sampled at the sub-Nyquist rate using a matrix Φ

in the sensing stage that is employed to capture a received waveform, where K is the number of data samples collected by the receiver, where i is a positive integer between 1 and N, indexing the columns of the dictionaries and where Φ

resembles an identity matrix I, where the value of any element along the diagonal line of the Φ

matrix is one and the value of any other element of the Φ

matrix is zero, except that only K samples (K<

<

N) are digitized from an input signal such that the rows of Φ

form a subset of the rows of the identity matrix I and the size of Φ

is (K×

N) as only K rows from the identity matrix I are selected.

5. The system of claim 4 wherein the selection of K rows from N rows of the identity matrix I could be derived from randomized selection or from any uniformly-spaced interval, which results in K digitized signal samples stored in a vector which lies directly on an original analog signal x(t).

6. The system of claim 5 wherein the matrix Φ

can be initialized once and be used among many records, or changed from one record to the next, but must be used in both signal sampling and signal reconstruction.

7. The system of claim 1 wherein the first and second dictionaries are coherent and the first dictionary comprises a matrix of digitized samples of a transmitted waveform at the Nyquist minimum sample rate stacked in column vectors, each column vector associated with an incremental delay, and wherein the incremental delay is substantially the highest increment possible that the system can support without an ineffective signal being produced.

8. The system of claim 1 wherein a linear combination for reconstruction of the signal is solved using an orthogonal matching pursuit technique wherein as α

is optimized such that y=Φ

x=Φ

Sα

, where α

is the sparse vector of weighting coefficients wherein significant coefficients reveal potential phase-shift and amplitude information of significant targets, and wherein the matrix Φ

is selected as a substantially random collection of rows of an identity matrix and samples of x are randomly and directly selected as measurements y.

9. The system of claim 1 wherein the first and second dictionaries are matrices which are constructed from time-shifted versions of a transmitted waveform of radar at two different sampling rates, the sampling rate for the first dictionary being above the Nyquist minimum sample rate and the sampling rate for the second dictionary being below the Nyquist minimum sample rate such that the columns of both matrices are constructed from digitized samples of the time-shifted transmitted waveform at different resolutions.

10. The system of claim 1 wherein a linear combination used for reconstruction of the signal comprises the use of time-shifted and weighted replicas of the received signal, where the weighting correlates to signal attenuation and the time shifting utilizes at least one parameter to model phase changes caused by different target properties and travel distances.

11. The system of claim 1 wherein the first and second dictionaries are adaptable to the transmitted data signals having different waveforms so as to be capable of storing time shifted versions of various different signal waveforms having a bandwidth range of approximately 300-3000 MHz, and wherein the at least one processor comprises a waveform reconstruction processor and wherein a real signal emitted from a radar transmitter is measured and employed in the construction of the first and second dictionaries.

12. The system of claim 1 wherein the at least one processor is a waveform reconstruction processor and the first dictionary can be represented by S where a received signal x(t) is sampled at the Nyquist minimum sample rate, the collected digitized data record can be represented as x=Sα

, where (N×

1) column vector x is a length-N record, the N×

(I×

J) matrix S is a redundant matrix whose columns contain the time-shifted versions of a transmitted signal s(t) and an (I×

J)×

1 column vector, and α

is the sparse vector of weighting coefficients and where (i, j) is defined as the coordinates of a pixel location of a reflective target in the recovery area with 1<

i<

I, 1<

i<

J.

13. The system of claim 1 wherein the first and second dictionaries comprise digitized signal vectors x and y, and wherein the digitized signal vector y is equal to Φ

times the vector x, where Φ

is a sampling matrix, and where the value of any element along the diagonal line of the matrix Φ

is one and the value of any other element is zero, and in the case where the number of samples collected by the receiver and digitized is K (K<

<

N), where N is equal to the samples obtained at the Nyquist minimum sample rate, the size of the matrix Φ

is K×

N and only K rows from the identity matrix are selected, and wherein the same matrix Φ

is used for signal sampling and signal reconstruction.

14. The system of claim 1 wherein the receiver comprises at least one radar sensor and wherein the analog to digital conversion system comprises a plurality of analog-to-digital converters each having an independent clock and operating in asynchronous fashion, each analog-to-digital converter operating on the output of the at least one radar sensor and collecting samples at uniform interval below a Nyquist minimum sample rate using the independent clock.

16. The method of claim 14 wherein the first dictionary can be represented by S where S is a redundant matrix whose columns contain the time-shifted versions of a transmitted signal s(t), and where the first dictionary is used to represent vectors x representing Nyquist length records and the second dictionary can be represented by S′

where S′

=Φ

S is used to represent sub-Nyquist vectors y and wherein the signal is reconstructed using the sparsity driven inverse problem

α

*=argmin|α

|_0,1 such that y=Φ

Sα

where α

is a sparse vector of weighting coefficients, α

* is the optimal α

, S is a redundant matrix whose columns contain the time-shifted versions of a transmitted signal s(t), Φ

is a sampling matrix, and wherein the sparsity driven inverse problem is solved using one of orthogonal matching pursuit, basis pursuit, gradient projection, matching pursuit, orthogonal matching pursuit, regularized matching pursuit or subspace pursuit.

17. The method of claim 16 wherein the first and second dictionaries are coherent and comprise vectors x and y, respectively, representing a plurality of time shifted responses representing a transmitted data signal;

the first dictionary comprising time shifted versions of transmitted signals which are sampled above the Nyquist minimum sample rate;

the second dictionary comprising time shifted versions of transmitted signals which are sampled below the Nyquist minimum sample rate and wherein the signal is reconstructed using at least one processor for reconstruction of the waveform signals;

the at least one processor operating to transform a sub-Nyquist digitized output into a real time data signal using the first and second dictionaries to produce a data signal as a linear combination of previously observed and weighed replicas of a transmitted data signal.

18. The method of claim 17 wherein reconstruction of a received signal comprises solving the L₀-norm optimization problem in the following equation

{circumflex over (α

)}_ij=arg min∥

α

_ij∥

₀s.t.y_ij=Φ

_ijSα

_ij

wherein the sparsest approximation possible of the vector records y_ij,data samples collected at the jth receiver containing echoes from a signal previously transmitted by the ith transmitter, is found as a linear combination of the corresponding vectors records x_ij represented by the first dictionary S, where α

_ij is a sparse vector of weighting coefficients, {circumflex over (α

)}_ij is the optimal α

_ij, Φ

_ij is a sampling matrix, and wherein the sparsity driven inverse problem is solved using one of orthogonal matching pursuit, basis pursuit, gradient projection, matching pursuit, orthogonal matching pursuit, regularized matching pursuit or subspace pursuit.

19. The method of claim 16 wherein the first and second dictionaries are coherent dictionaries, the first dictionary comprising vectors representing a received data record x as x=Sα

where x represents column vectors containing Nyquist length records and matrix S is a redundant matrix whose columns contain time shifted versions of the transmitted signal and α

is the sparse vector of weighting coefficients wherein significant coefficients reveal potential phase-shift and amplitude information of significant targets, and wherein the second dictionary is a sub-Nyquist dictionary defined as y=S′

α

, where y is the sub-Nyquist version of x and S′

=Φ

S is the sub-Nyquist version of S, and Φ

is a sampling matrix and y=Φ

x.

20. The method of claim 16 wherein the reconstruction of a received data signal further comprises using linear combination of previously recorded transmitted signals for signal recovery by using an orthogonal matching pursuit technique wherein α

is optimized such that y=Φ

x=Φ

Sα and

wherein the matrix Φ

is selected as a substantially random collection of rows of an identity matrix I and samples of x are randomly and directly selected as measurements of y.

15. A method for reconstructing data signals from one of incomplete, compressed or sparse measurements comprising:

providing a receiver for receiving data signals from a target area,

providing an analog to digital conversion system operatively connected to the receiver, the analog to digital conversion system digitizing a received data signal at a slower rate than a Nyquist minimum sampling rate to obtain a sub-Nyquist digitized output,

creating first and second dictionaries for storage of a plurality of time shifted data recovered from previously transmitted signals;

the first dictionary operating to store time shifted versions of previously transmitted signals which are sampled above the Nyquist minimum sample rate;

the second dictionary operating to store time shifted versions of previously transmitted signals which are sampled below the Nyquist minimum sample rate,

providing at least one processor for reconstruction of waveform signals;

the at least one processor operating to transform the sub-Nyquist digitized output from the analog to digital conversion system into an approximation of the received data signal using the first and second dictionaries to produce the data signal.