Multi-sparse signal recovery for compressive sensing.

Area of Science:

• Signal Processing
• Optimization Theory
• Biomedical Engineering

Background:

• Compressive sensing (CS) enables signal recovery from sub-Nyquist measurements.
• Traditional CS methods rely on sparsity in a single domain, often using L0 norm optimization.
• Some signals exhibit sparsity across multiple domains, a property not fully leveraged by classical techniques.

Purpose of the Study:

• To develop an improved signal recovery method for signals with multi-sparsity.
• To introduce a novel convex programming model that incorporates multiple sparsity constraints.
• To enhance signal reconstruction performance by utilizing additional prior information from multi-domain sparsity.

Main Methods:

• Formulation of a new convex programming model incorporating multiple sparsity constraints across different domains.
• Inclusion of a linear measurement fitting constraint within the optimization framework.
• Numerical experiments using electromyography (EMG) signals, known to possess sparsity in both time and frequency domains.

Main Results:

• The proposed method demonstrates superior signal recovery performance compared to classical approaches for multi-sparse signals.
• Exploiting sparsity in multiple domains provides a significant advantage in reconstruction accuracy.
• EMG signal reconstruction showed improved results when applying the multi-sparsity convex programming model.

Conclusions:

• The novel convex programming approach effectively leverages multi-domain sparsity for enhanced signal recovery.
• This method offers a significant improvement over traditional single-domain sparsity techniques.
• The proposed technique shows promise for applications involving signals with inherent multi-sparsity, such as EMG analysis.