Reach On Blind Source Separation In Wireless Sensor Networks

As the important component of the perception layer of the Internet of Things, wireless sensor networks have been widely used for signal detection, estimation, target tracking, and so on. In actual scene, multiple sources may appear simultaneously. In order to acquire the signal of interest and remove interference from others, it is necessary to separate the mixed signals collected by the sensors for subsequent process and application. Wireless sensor networks consist of numerous sensors which are randomly deployed and self-organized. And it is hard to get the priori of the multiple signals and how they’ve mixed. That is, in wireless sensor networks, the separation of the mixed signals is blind. Thus, blind source separation in wireless sensor networks is a crucial issue in the field of signal processing in wireless sensor networks.Taking the energy limit of wireless sensor network in to consideration, a kind of method was proposed for signal number estimation in wireless sensor networks, which consisted of clustering, judging within clusters and decision fusion. Meanwhile, a new cluster head selection algorithm was proposed base on LEACH, in which theresidual energy of sensors was taken into account when selecting new cluster heads. In the method for signal number estimation, the wireless sensor network was divided into multiple clusters, and then each cluster estimated the number of source independently, and each cluster head sent the estimation result to the sink node. Finally the sink node fused the result from all the clusters to work out the final conclusion. Simulations showed that the method for estimation of the number of signals and the algorithm for cluster head selection we proposed could reduce the energy consumption for transmission, extend the lifetime of the network, and achieve high accuracy at the same time.Accurate separation of signals emitted by different targets is an essential precondition for multi-target tracking in wireless sensor networks. A sensor selection scheme for blind source separation in wireless sensor networks was proposed. Considering the sensors’ observation signal-to-noise-ratios and their residual energy, we first fitted the mean observation signal-to-noise-ratio of sensors and the performance of blind source separation; and then modeled sensor selection as a combinatorial problem. The sensors to be selected were determined through solving the combinatorial problem by a heuristic method based on convex optimization. Simulations showed that the sensor selection we proposed was capable to increase the residual energy of sensors, as well as guarantee the high and stable performance of blind source separation.Lastly, blind source separation with noise and transmission error in wireless sensor networks was studied. The effect of transmission error on blind source separation was examined. Then a method for outlier detection based on clustering and distance was proposed, which was used for detecting and modifying the outliers in the recovery signals caused by transmission error. Simulations showed that the methodwas capable to reduce the effect of transmission error and increase performance of blind source separation.