From Theory to Practice: Sub-Nyquist Sampling of Sparse Wideband Analog Signals
read more
Citations
Block-Sparse Signals: Uncertainty Relations and Efficient Recovery
Beyond Nyquist: Efficient Sampling of Sparse Bandlimited Signals
Structured Compressed Sensing: From Theory to Applications
Spectrum Sensing for Cognitive Radio : State-of-the-Art and Recent Advances
Robust Recovery of Signals From a Structured Union of Subspaces
References
Compressed sensing
Robust uncertainty principles: exact signal reconstruction from highly incomplete frequency information
Atomic Decomposition by Basis Pursuit
The restricted isometry property and its implications for compressed sensing
Optimally sparse representation in general (nonorthogonal) dictionaries via 1 minimization
Related Papers (5)
Frequently Asked Questions (14)
Q2. What future works have the authors mentioned in the paper "From theory to practice: sub-nyquist sampling of sparse wideband analog signals" ?
The engineering aspects are the prime focus of the current paper, while future work will sharpen the theoretical understandings and report on circuit-level implementation [ 12 ].
Q3. What is the key for reducing recovery complexity?
The CS problem of the CTF, (19), is used only for support recovery, which is the key for reducing recovery complexity and allowing low-rate processing.
Q4. What is the signal model for which the random demodulator was de-signed?
The signal model for which the random demodulator was de-signed consists of multitone functions:(40)where is a finite set of tones(41)The analysis in [10] shows that can be recovered from , using the linear system(42)where is matrix and collects the coefficients .
Q5. How many channels are used for correct recovery?
The results show that in the high SNR regime correct recovery is accomplished when using channels, which amounts to less than 18% of the Nyquist rate.
Q6. What is the noise space associated with the remaining negligible eigenvalues?
The noise space, which is associated with the remaining negligible eigenvalues is discarded by simple thresholding ( is used in the simulations).
Q7. What is the CS algorithm required to run on a huge scale system?
In turn, the CS algorithm would have to run on a huge-scale system, dictated by the ambient Nyquist dimension, which is time and memory consuming.
Q8. What is the ability to activate the digital recovery of any specific band of interest?
The controller has the ability to selectively activate the digital recovery of any specific band of interest, and in particular to produce an analog counterpart (at baseband) by overriding the relevant carrier frequencies.
Q9. How long does the CTF have to wait for a new support estimate?
In a real-time environment, about consecutive input vectors should be stored in memory, so that in case of a support change the CTF has enough time to provide a new support estimate before the recovery of , (29), reaches the point that this information is needed.
Q10. What are the practical considerations for the implementation of the analog front-end?
the authors presented various practical considerations, both for the implementation of the analog front-end (e.g., setting the number of channels, trading system branches by a higher sampling rate, and some potential hardware simplifications), and for the digital stage (e.g., low-rate and real-time processing, handling time-varying spectrum, and quantization).
Q11. How long does the lowrate sequences remain valid?
As a result, whenever the spectral support changes, the lowrate sequences remain valid only for 20 cycles, and then becomes invalid for 30 more cycles, until the CTF provides a new support estimate.
Q12. What is the sign pattern of m-sequence?
In practice this is not an obstacle, since standard logic gates and feedback can be used to generate a sign pattern of length (a.k.a, m-sequence) with just a few components [19], [20].
Q13. How can the authors detect the change in the spectral support?
In order to notice the support changes once they occur, the authors can either rely an indication from the application layer, or automatically identify the spectral variation in the sequences .
Q14. What is the advantage of the current mixing stage?
In contrast, the current mixing stage has the advantage that the IMV is expressed directly in terms of the low-rate sequences , and the computation of in (28) is carried out directly on the input sequences.