Study On Human Movement Monitoring Oriented Signal Compressing And Activity Recognition In Wireless Body Sensor Networks

Wireless body sensor network as a new miniarure wireless sensor network technology has attracted wide attention. It is a kind of emerging technology with great practical applicability, including health care, life and entertainment, military communications, emergency and disaster relief, and aerospace etc. The human activity monitoring system based on wireless body sensor networks is one of the hot issues in this field. This paper revolves around the practical problems of wireless body sensor networks existing in the process of human activity recognition. This paper has a strong focus on the problems of signal denoising, feature extraction, classification method and data compression. The main work and achievements are as follows:1. A new threshold function is proposed to overcome the disadvantage of traditional wavelet threshold denoising method. It overcomes the fixed attenuation of soft-threshold method and the oscillation of hard-threshold method. For some activities with high similarity, this paper proposes novel feature extraction method. And this method effectively improves the efficiency of classification.2. In view of problem of the traditional support vector machine (SVM) in the optimal parameter selection, the particle swarm optimization (PSO) algorithm is used to optimize SVM. The experimental results show the superiority of this method than the traditional parameter selection method. And it obtains a higher recognition rate than the traditional BP neural network and KNN classifier.3. Block Sparse Bayesian Learning (BSBL) is proposed as a new method to the data compression problem of wireless body sensor networks. It has a very good compression and reconstruction effect on the not strictly sparse acceleration signal. The comparison of the traditional compress sensing algorithm Basis Pursuit and COSAMP shows the superiority of this algorithm.4. The sparse representation classifier is used to solve the data compression and activity recognition problem in wireless body sensor network at the same time. In view of the computational complexity problem of traditional sparse representation classifier, a means of replacing the local base with contiguity class is proposed.5. An elderly daily behavior monitoring system based on wireless body sensor networks is designed. It can effectively identify the common human activities, especially the fall action.All in all, the study on human movement monitoring oriented signal compressing and activity recognition in wireless body sensor network has great theoretical significance and practical application value. It is worth for intensive research.