Sensor-based Fog-Cloud Integrated Human Fall Detection System using Regression-based Gait Pattern Recognition

Human falls areidentified as the common problems among the elderly persons living lonely in home may lead to severe injuries, continuing disabilities, fear of falling and even death. The goal of this research study is to develop the wearable sensor-based system to assess the human fall risk and gait impairments of elderly persons in daily living conditions. So, a fog-cloud integrated human fall detection system (FCIHFDS) is proposed to collect the gait parameters for identifying the fall risk of remote patients. In order to support real-time learning and recognition of gait patterns, a multiple hidden Markov model-based regression learning (MHMMRL) classifier is introduced to deal with on-board human fall and other behavioral activities prediction. As a result, the proposed FCIHFDS provides better patient activities recognition against the existing fall prediction systems in terms of both prediction time and prediction accuracy. To minimize the latency and cost involved during the prediction and rehabilitation process, a novel cost aware offloading scheme is introduced for appropriate decision making in the peer-to-peer fog-cloud integrated platform.The proposed cost aware offloading scheme outperform the existing Energy and Time Efficient Offloading, and Cost and Energy Aware Offloading schemes in terms of cost(utility value). These findings may be more useful for supporting long-time monitoring of human fall risk at ambient assisted living environments and improve its clinical utility.