The Research And Applications Based On Distributed Compressive Sensing

Traditional signal processing is based on Nyquist sampling theorem, causing huge waste of storage space because of large amounts of data. Compressed sensing (CS) is a new sampling theory and has been proved that it is possible to reconstruction signals, images and other types of data exactly using less than the Nyquist sampling rate. CS is an emerging field based on the revelation that a small collection of linear projections of a sparse signal contains enough information for reconstruction. Furthermore, distributed compressed sensing (DCS) exploits the correlation among multiple signals to achieve further compression.The huge demand for information results in enormous pressure of signal sampling, transport and storage. How to relieve this pressure and how to extract the useful information of signal effectively are problems needed to solve. The appearance of CS provides a solution to alleviate these problems. In addition, due to the growing of network, the application of DCS has a very wide range of prospect; it also gradually entered into the practical field.In this paper, the history of CS is referred, and we introduce some key technologies in detail, including sparse representation-of signal, observation matrix and reconstruction algorithm design. By some inductions, the related problems and the current researches are concluded.For video/image signal scenarios, by applying CS, the storage efficiency and data compressibility can be significantly improved. DCVS can obtain higher compression ratio than CS with the usage of intra-frame and inter-frame correlations among the frames of video sequence. In addition, DCVS also shifts the complexity burden from encoder to decoder, resulting in low complexity video coding.The reason why the correlation model can improve the efficiency of coding is revealed, and the particularity of reconstruction algorithm is inferred. Moreover, based on sparse filter correlation model (SFCM) we exploit the correlations among successive video frames under the framework of DCVS. At the central decoder, joint reconstruction is accomplished with the assistance of modified belief propagation (BP) algorithm Simulation results illustrate that the proposed method provides better PSNR performance than joint sparse model 1 (JSM1) for DCVS.