A Dynamic Bayesian Recovery Algorithm for Time Series Signals from Compressive Measurements

In this paper, a dynamic sparse Bayesian learning algorithm is proposed for reconstruction of time series signals from compressive measurements. The proposed algorithm introduces correlation matrix to protect the inner correlation among signals, and recursively estimate interested parameters via online learning, which further introduces correlation between adjacent measurements batches. Experimental results showed that the proposed algorithm has much higher success rates and signal-to-error ratio than that of the traditional sparse Bayesian learning algorithm and Block-sparse Bayesian learning algorithm.