Compressed Sensing via Dictionary Learning and Approximate Message Passing for Multimedia Internet of Things

In this paper, we present a compressed sensing-based approach, which combines the dictionary learning (DL) method and the approximate message passing (AMP) approach. The approach can be used for efficient communication in the multimedia Internet of Things (IoT). AMP is a signal reconstruction algorithm framework, which can be explained as an iterative denoising process. On the other hand, the DL method seeks an adaptive dictionary for realizing sparse signal representations, and provides good performance in signal denoising. We apply the DL-based denoising method within the AMP algorithm framework and propose a novel DL-AMP framework. We demonstrate our framework's effectiveness for multimedia IoT devices by showing its capability in reducing required communication bandwidth for multimedia communication while improving reconstruction quality (by over 2 dB).