Human Movement Classification using Signal Level Fluctuation in WBAN at 403.5 MHz and 2.45 GHz

摘要

There are many requirements for Body Area Network (BAN) system. BAN system needs to be reliable since it is often used for carrying current body status of patients or elderly people. One challenge of BAN is that the reliability of the system is suffered from intentional movements and unintentional movements. Therefore, the human movement should be considered in order to maintain the quality of BAN system. Various movements cause signal fluctuation differently, so the human movement could be determined by analyzing the received radio signal level. This paper aims to present that the human movement can be recognized by a temporal received signal level. A neural network is applied as a classification tool. The human movement data is based on a numerical simulation generated in 2 frequency bands, which are 403.5 MHz and 2.45 GHz. The result shows that the neural network using the received signal level form six sensors concurrently can classify the human movement well. The accuracy measured by test set data is around 90 percent. However, the accuracy reduces to around 66 percent when the neural network uses the received signal level form one sensor.