Strategy Of Enhancing Energy Efficiency In Wearable Computing

With the rapid development of micro-electro machenical system(MEMS), micro controller unit(MCU) and wireless communication technology, the motion capture system based on inertial measurement unit(IMU) presents tremendous value of research and broad application prospects. This kind of IMU has a number of advantages such as low-cost, small-sized and portable, which might be widely applied into the fields such as elder-monitoring system, rehabilitation system, human computer interaction, and so on. However, the requirements of low power consumption and real-time communications at high rate constrain its further application. Monitoring in long duration and real-time stream transmitting in high speed may result in poorly energy efficient system.Therefore, this thesis explores the strategies of improving the energy efficiency of wearable computing for human motion tracking. The main research work includes:(1) Proposing an extended kalman filter algorithm with mitigating the extra linear acceleration.An extended kalman filter is designed in this paper to estimate the orientation of the rigid body, which is able to eliminate the process error and measurement error of inertial measurement units. Besides, an adaptive error covariance matrix is presented to reduce the influence of extra linear acceleration in spinning motion. The experimental results show that the root mean square error of extended kalman filter proposed in this thesis is 0.7 degree.(2) Implementing the online orientation estimation algorithm to improve the energy efficiency.This thesis implements the extended kalman filter algorithm for the wearable platform limited by computing resources, which can decrease the quantity of data transmission, reduce energy consumption of nodes and extend the network lifetime. During the implementation, application components, sensor components and operating components have been designed to satisfy the requirement of high accuracy. In addition, the data blocking cache mechanism and optimization of trigonometric function computation are proposed to guarantee real-time data processing. Experimental results show that the online orientation estimation algorithm can prolong the life of system by 20.6%.(3) Proposing an adaptive data fusion algorithm to further improve energy efficiency.To combat the problem of high power consumption due to high rate transmissions, this thesis proposes an adaptive data fusion algorithm which extracts the key frames of human motion in real time. The sensor node only sends these key frames. And on the base station, the human motion is predicted and reconstructed. Experiment results show that this algorithm can further reduce data transmissions, improve the energy efficiency significantly and prolong the system life by 76.7%.