Research On Energy Efficient Motion Recognition Algorithm In Wireless Body Area Networks

Wireless body area network(WSN)is a kind of wireless sensor network,which is composed of a portable,wearable or implantable sensor node.The wearable devices provide a new way for human health monitoring,which has a huge demand and practical significance in the application of human health and disease monitoring.Human action recognition based on inertial sensors is widely used in human motion monitoring.By analyzing the sensor data collected from different parts of the human body,we can analysis the state of human life.In human motions recognition based on the inertial sensor in wireless body area network,it is a very challenging problem that how to exploit the temporal and spatial correlation of sensor data to meet requirements of human activities and how to make better use of energy and improve the effectiveness of energy for sensor nodes which is constrained by energy.In this paper,the wireless body sensor network with several wearable inertial sensors is taken as the research object.And the temporal and spatial correlation of sensor data,adaptive compressed sensing from action recognition is studied through sparse Bayesian learning framework and compressed sensing theory,which contains data compression,data fusion,effective energy.The main work includes the following points:Initially,an action recognition method based on spatial and temporal sparse Bayesian learning framework and called STSRC-BO is proposed.According to the existing temporal and spatial correlation between motion data in wireless body area network,we use Spatial temporal sparse Bayesian learning framework,which can exploit the temporal and spatial correlation of sensor data to compress more data,reduce the transmission energy consumption.The experimental results show that the STSRC-BO method has high recognition rate more than other methods when the compression ratio is gradually increasing.At the same time,with the increase of compression ratio,the STSRC-BO method can maintain a relatively stable recognition rate.Furthermore,an adaptive compressed sensing mechanism based on different recognition rate feedback is proposed.Because the sensor data in different body positions are related and correlation of different motion data will be different,so we get different recognition rate under different compression ratio for different actions.In order to make use of the different recognition accuracy of different activities under different compression ratios,an adaptive compression scheme based on recognition rate feedback is proposed.The experimental results show the adaptive mechanism can reduce the energy consumption of data transmission compared with non-adaptive mechanism.Finally,in order to verify the performance of the algorithm,a prototype system is designed and implemented.The prototype system is used to collect the motion signals in the daily life.