Human Multi-gesture Passive Recognition And Localization Technology Using Wireless Sensor

Device-free passive localization technology based on wireless sensor has attracted much attention as it does not require the subjects to carry wearable devices and can effectively protect the privacy of users. At present, many kinds of these localization technologies mainly deploy single-layer wireless links structure to locate the target position. However, these technologies cannot effectively identity gesture features of the target. It is obvious that these technologies could not do anything about determine the position of old people fallen down and influence the elderly rescue in the process of elderly care. But fall detection technology mostly need wearable devices and these means obviously do have service restriction and poor practicability.By combining with relative theories, this dissertation mainly makes an analysis around target location and recognition of human gesture, then proposed a device-free passive recognition of human gesture and localization technology based on wireless sensor in indoor environments. The technology we proposed can effectively detect target position and gesture information with a joint recognition model by building two-layer wireless links. Apparently our method has certain application prospects.The main contributions in the dissertation can be summarized as follow:(1) Data acquisition technology research based on state switching, can availably split the original data stream of different states and improve the precision of the data set. In addition, this dissertation advanced a data preprocessing solution using sliding window, to resolve the problem of loss packet in the data acquisition process. This solution could fill with the raw data packet loss to improve real-time utilization.(2) We propose a human behavior recognition model using two-layer wireless links and make an experimental analysis of the model. To verify the applicability of the model, this dissertation launched experiments in a real indoor environments and use a variety of multiple classifiers to contrastively analyze. Experiments show that the proposed model can effectively detect four simple behaviors:normal environment, human walking, standing and lying down.(3) We further propose a human multi-gesture localization technology in cluttered indoor environments which depend on a method named location and human gesture fingerprinting. This dissertation did the research about two position gestures:human standing and lying down, then respectively chosen three different types of indoor environments to test the robustness of the above method. Besides that, this dissertation also studied human multi-gesture tracking. Experimental results show that the average recognition accuracy of the proposed method in three environments can reach 91.91%.