TY - GEN
T1 - Prediction of received signal strength from human joint angles in body area networks
AU - Tran, Thang Manh
AU - Vejarano, Gustavo
PY - 2016/2/1
Y1 - 2016/2/1
N2 - Focusing on movements of a human participant performing physical-therapy exercises, this paper presents an algorithm that predicts the received signal strength indicator (RSSI) of wireless sensor nodes attached to the user. The body area network (BAN) formed by the nodes is a motion capture system that measures joint angles of the user at the shoulder and elbow. In order to predict the RSSI, we first show that the wireless signal experiences severe attenuation from human-body shadowing even though distances between transmitters and receiver are less than 3 meters. Second, we show that the RSSI fluctuates periodically with regular body movements (i.e., physical-therapy exercises). We then model the movements using k-means clustering and Markov chains and determine the probability distribution of the RSSI at each state in the movement. Finally, the RSSI is predicted with a maximum a posteriori probability (MAP) detector. Experimental results show that the RSSI can be predicted with a root mean square error (RMSE) of 3.7 dB, which is an error within 4.2% of the average RSSI level, and when a prediction is made, it is valid for the next 1083 milliseconds (ms) on average.
AB - Focusing on movements of a human participant performing physical-therapy exercises, this paper presents an algorithm that predicts the received signal strength indicator (RSSI) of wireless sensor nodes attached to the user. The body area network (BAN) formed by the nodes is a motion capture system that measures joint angles of the user at the shoulder and elbow. In order to predict the RSSI, we first show that the wireless signal experiences severe attenuation from human-body shadowing even though distances between transmitters and receiver are less than 3 meters. Second, we show that the RSSI fluctuates periodically with regular body movements (i.e., physical-therapy exercises). We then model the movements using k-means clustering and Markov chains and determine the probability distribution of the RSSI at each state in the movement. Finally, the RSSI is predicted with a maximum a posteriori probability (MAP) detector. Experimental results show that the RSSI can be predicted with a root mean square error (RMSE) of 3.7 dB, which is an error within 4.2% of the average RSSI level, and when a prediction is made, it is valid for the next 1083 milliseconds (ms) on average.
UR - https://digitalcommons.lmu.edu/cs_fac/39/
U2 - 10.1109/ICCNC.2016.7440700
DO - 10.1109/ICCNC.2016.7440700
M3 - Conference contribution
T3 - 2016 International Conference on Computing, Networking and Communications (ICNC)
SP - 1
EP - 6
BT - Prediction of received signal strength from human joint angles in body area networks
PB - IEEE
ER -