The human movement identification using the radio signal strength in WBAN

This paper investigated the feasibility of using the radio signal strength of sensors placed around the human body in the human movement identification. This proposed method can identify the human movement in WBAN using only the radio signal strength, thus any additional tools are not necessary. OpenNICTA provides the BAN measurement channel in three kinds of human motions, which are running, walking and standing. This paper used three sets of the measurement data, which Tx-Rx located at Back-Chest, RightAnkle-Chest, and RightWrist-Chest. Each data set was separately used to identify the movements. This paper used two types of machine learning, which are neural network and decision tree. In the neural network, it has been found that using eight types of features, which are SCP, Range, SSI, RMS, LCR, SC, WAMP, Histogram, calculated from 200 continuous received signal levels can identify the human movements with accuracy rate of 90.41-98.83 percent. Using the same features, the decision tree can identify the human movements with the accuracy rate of 99.04-99.66 percent. Both tools perform well on the human movement identification. However, the decision tree outperforms the neural network in this task.