Study On Compressive Sensing Based On Uncertain Environment In IOT

To connect things, a variety of sense components and equipments are adopted to gather and exchange information in real time and are combined with Internet as a huge network, which is called Internet of Things(IOT). In this huge network, the wirless technology is critical. In IOT, it is crucial to be sensed information of uncertain environment reliably and transmit them effectively. Recently, a novel signal sampling theory Compressive Sensing is proposed, which only acquires few samplings from sparse signals and then could recover them precisely. Therefore, it has been paid attention and applied in many fieds. Based on compressive sensing, sensors only need to transmission a small number of samplings to the termination, and then the receiver can recover the original signals accurately, which is an advantage for wireless sensor network.However, it is difficult for uncertain environment information in IOT to be satified the requirement of sparsity of compressive sensing. Therefore, the characters of the sensed siginals should be further analyzed to obtain the compresiable content, and then design suitable approaches. Due to the diffucult of sparsing signals, this paper studies on signals which are appeared in three scenes in IOT and then design corresponding methords. The main contents in the paper are as follow:Based on the scene that it is difficult for relative signals to obtain their sparsity, a compresensive sensing approach based on local regions has been proposed. The structure of this kind of data consists of two parts, which are common sparse part and special sparse part respectively. In this case, the former should be abstracted as much as possible. In this paper, two reasonable assumptions are given and then a spatical-temporal correlation model is presented. In this model, data gathered in local region satisfy spatical-temporal correlation in a certain extend. To abstract their common parts, compressive matrices are designed for four data distributions, which are uniform distribution, Guassian distribution, expential distribution and Poission distribution. For the two former ones, hybrid compressive matrices are proposed. For the two later ones, the functions of transformation are proved and the paramaters estimation could be easily achieved.Based on the scene of multi-classifications signals, a compressive sensing method based on classification is proposed. The effective hybrid transmition schem is present, in which both the advantage of compressive sening is retained and the unnecessary energy cost is avoided. To increase the accurate of reconstructing signals, the choice of measurement matrices and the number of classes are considered. The classification-based algorithm is designed and the principle of sampling is presented to decrease required samplings and redundant samplings.Based on the scene of random signals without sparsity, a compressive sensing approach for limited random signals is proposed. Due to the limitation of the sensing ability of equipments, the gathered data are limited.. In this limited range, random signals are encoded for sparsity, which turn this kind of signals into sparse ones in a uniform formate to satisfy the requirement of theory of compressive sensing. To satisfy the requeirement of CS, binary digits are sparsing encoded. Meanwhile, suitably layering binary digits to decrease the spase of matching and increase the success of pursuiting. Afterwards, the reconstruction algorithm based on layer is designed to reduce the requirement of samplings.Our contributions in the paper are as follows:1. The CS approach based on local region data abstract the common part from the data which immensely the required number of samplings and the process of compressing signal need rarely energy cost.2. The CS approach based on classification reduces the samplings and avoids redundant ones. Meanwhile, the hybrid scheme decreases the number of signal transmission and guarantees the load balance.4. The CS approach based on limited random signals transforms random signas to sparase signals with an uniform format which satisfies the requriemnt of CS and is in favor of diminish samplings, and the lengh of package is stable in the process of transmission.