Compressive Sensing
Compressive Sensing (Compressed Sensing) Signal Acquisition · Also known as: Compressed Sensing, CS, Sparse Recovery, Sub-Nyquist Sampling
Compressive Sensing (CS) is a signal acquisition and reconstruction technique that exploits signal sparsity to recover high-resolution signals from far fewer samples than required by the Nyquist sampling theorem. Developed by Emmanuel Candès, Justin Romberg, and Terence Tao in 2006, compressive sensing challenges the traditional sampling paradigm by showing that signals with sparse representations can be reconstructed from sub-Nyquist random measurements using nonlinear optimization.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use compressive sensing when signals are sparse in known bases and acquisition costs (time, energy, radiation) are high, such as medical imaging, radar, and remote sensing. It is ideal when measurement hardware is expensive or power-limited. Avoid compressive sensing when signals are not sparse or when noise is high relative to signals—recovery quality depends on sparsity and measurement signal-to-noise ratio.
Strengths & limitations
- Enables sub-Nyquist sampling and acquisition, reducing data acquisition costs, time, and power consumption
- Provides exact signal reconstruction with extremely few measurements when sparsity condition is met
- Random measurement matrices are universal—same measurements work for different sparse bases
- Naturally robust to missing or corrupted measurements through convex optimization formulation
- Theoretical guarantees on recovery success provided through restricted isometry property analysis
- Requires signals to be sparse in some known basis; non-sparse signals cannot be recovered from few measurements
- Reconstruction requires solving nonlinear optimization problems, computationally expensive compared to simple sampling
- Success depends on proper choice of sparse basis; recovery fails if sparsity basis is wrong
- Requires incoherence between measurement basis and sparsity basis, not always achievable in practice
- Performance degrades significantly in high noise regimes where measurement SNR is low
Frequently asked
What is the relationship between sparsity and the number of measurements needed?
For a k-sparse signal (k nonzero coefficients), recovery requires at least O(k·log(N)) random measurements, where N is the full signal dimension. This is far fewer than the N measurements required by Nyquist sampling. The required measurement count depends on sparsity level k and signal dimension N. Higher sparsity requires proportionally fewer measurements, but very high sparsity is needed for extreme sub-Nyquist factors.
What is the restricted isometry property and why is it important?
The restricted isometry property (RIP) measures how well a measurement matrix preserves distances between sparse signals. A measurement matrix satisfies RIP-k if it approximately preserves the norm of all k-sparse vectors. Strong RIP guarantees successful recovery of all k-sparse signals. RIP depends on both the measurement matrix and the sparsity basis; proper design of incoherent measurement and sparsity bases ensures good RIP.
How do I choose the sparse basis for my signal?
Ideally, choose a basis where your signal has few nonzero coefficients. Common choices are Fourier (for sinusoidal content), wavelets (for transient and multi-scale content), or learned dictionaries from training data. If unsure, try multiple bases and verify sparsity: examine histogram of coefficients and check that most are near zero. The sparser the signal in your chosen basis, the better compressive sensing will work.
What optimization algorithms are used for signal recovery?
The most rigorous approach is basis pursuit (BP), solving L1-minimization: minimize ||x||₁ subject to Φx = y. BP has strong theoretical guarantees but is computationally expensive. Faster alternatives include iterative thresholding (ISTA/FISTA), orthogonal matching pursuit (OMP), and stagewise algorithms. For real-time applications, greedy algorithms like OMP provide fast approximate recovery. Trade-off exists between computational speed and recovery quality.
Sources
- Candes, E. J., Romberg, J., & Tao, T. (2006). Robust Uncertainty Principles: Exact Signal Reconstruction from Highly Incomplete and Inaccurate Measurements. IEEE Transactions on Information Theory, 52(2), 489–509. DOI: 10.1109/TIT.2005.862083 ↗
- Eldar, Y. C., & Kutyniok, G. (2012). Compressed Sensing: Theory and Applications. Cambridge University Press. link ↗
How to cite this page
ScholarGate. (2026, June 3). Compressive Sensing (Compressed Sensing) Signal Acquisition. ScholarGate. https://scholargate.app/en/signal-processing/compressive-sensing
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Adaptive LMS FilterSignal Processing↔ compare
- FIR Filter DesignSignal Processing↔ compare
- Power Spectral Density EstimationSignal Processing↔ compare
- Short-Time Fourier TransformSignal Processing↔ compare