A Machine Learning Based Method for Coexistence State Prediction in Multiple Wireless Body Area Networks

The coexistence problem occurs when a wireless body area network (WBAN) is close to other WBANs. The degraded communication performance may even threaten people’s life for medical applications. Therefore, a reliable method for coexistence state prediction is required to ensure this problem could be detected and handled in time. This chapter presents a machine learning based method for coexistence state prediction, in which decision tree (DT) and naive Bayes classifier (NBC) are considered as supervised learning methods. Firstly, the average packet error rate (PER) and other 11 features extracted from signal to interference plus noise ratio (SINR), which have less computational cost, are selected and improved to reflect human relative movement and interference strength for feature extraction. Then, a set of models based on DT and NBC are generated with various configurations of different features, the number of continuous SINRs and the decision function. Finally, all models are compared and the best one is selected for state prediction. In the experiment, the data for training and testing is collected by CC2530 2.4 GHz low-power transceivers. Simulation results show that, compared with NBC based method, the proposed method is better in terms of accuracy and timeliness.