DMP-IOT: A distributed movement prediction scheme for IOT health-care applications

We have used second-order HMM for mobility prediction in mobile IP-based WSNs. To our knowledge, no learning method has been applied for modeling movement in this context.We have proposed a novel cell-based tree like structure by which we efficiently distribute the movement prediction model. This scheme makes online movement prediction efficient, while improving the handoff cost.The proposed scheme has a novel recovery mechanism which is invoked during false prediction of patients movement. Thus, the accuracy of prediction is not degraded. Display Omitted Mobility prediction in IP-based WSNs makes it possible to predict the next movement direction of mobile sensor nodes, which results in less power consumption and delay during handoff. The previous direction detection approaches need specific hardware facilities and impose considerable overhead during handoff. The advantage of DMP-IOT is distributing the learning model data around static sensors of the proposed tree, after the training phase. DMP-IOT includes a recovery mechanism that avoids disconnection of the mobile sensor node(s) to the network in case of a false prediction. The simulation results show about 25% improvement of DMP-IOT in saving power consumption and reducing handoff delay and packet loss, compared to movement direction approaches in similar works. The accuracy of the proposed movement prediction scheme is 83%, in average, which is validated by tStudent statistical test. Comparing the second-order Hidden Markov Model (HMM) with ANN reveals the superiority of the second-order HMM model in our application.