adaptive non-uniform compressive sampling (ANCS) of time-varying signals, which are sparse in a proper basis, is introduced. ANCS employs the measurements
of previous time steps to design the measurement matrix and distribute the sensing energy among the coefficients more intelligently.
In many applications, coefficients of the signal of
interest have different importance levels and the region of interest (ROI) is not known a priori. For instance, the salient area in a sequence of video frames or support of a sparse signal can be considered as the ROI. Our proposed method employs
the estimation of signal at each time step to approximate the importance level of each coefficient. This information is then exploited to devise a measurement matrix for the next time
step so that the reconstruction error of the more important coefficients are reduced significantly. To this aim, a Bayesian inference method is proposed that does not require any prior knowledge on importance level of coefficients, their probability
distributions, or sparsity of the signal. The proposed method is universal in the sense that it does not depend on the recovery method, sparsity of the signal, or the ROI detection method.