Complex multitask compressive sensing using Laplace priors

Abstract

Abstract Most existing Bayesian compressive sensing (BCS) algorithms are developed in real numbers. This results in many difficulties in applying BCS to solve complex‐valued problems. To overcome this limitation, this letter extends the existing real‐valued BCS framework to the complex‐valued BCS framework. Within this framework, the multitask learning setting, where L tasks are statistically interrelated and share the same prior, is considered. It is verified by numerical examples that the developed complex multitask compressive sensing (CMCS) algorithm is more accurate and effective than the existing algorithms for the complex sparse signal reconstructions

Keywords

• Mathematical analysis
• Signal processing and detection
• Signal processing theory
• Other topics in statistics
• Mathematical analysis
• Other topics in statistics