AS-LR Emergency Detection Scheme for Biomedical Applications

The fast development of medical sensors allowed the emergence of a contemporary biomedical application called Body Sensor Network (BSN). BSN allows medical personnel to monitor the status of the person, through analyzing physiological parameters (Heart Rate (HR), Blood Pressure (BP), body temperature (To), etc.) collected by miniature medical sensors placed on the human body. The main aim of BSN is to enhance people's life through detecting emergency cases as fast as possible and taking proper actions before it is too late. Since sensors in BSN have very limited energy that should be preserved, the challenge arises in designing schemes that ensure correct and fast emergency detection while maintaining low energy consumption of nodes. Machine Learning (ML) techniques are vastly adopted nowadays for anomaly detection due to the large amount of data collected by the sensors. Even though these techniques provide high percentage of accuracy in detecting emergencies, they operate over the whole dataset which quickly drains the energy of nodes due to the large amount of data that should be transmitted. Thus, in this article, we propose an algorithm that efficiently decreases the amount of transmitted data through dynamically adapting the sampling rate of every sensor based on the risk of emergency detected by the Linear Regression (LR) ML model. Adaptive sampling is used to reduce the energy consumption of nodes and to accelerate the detection of emergencies, while LR is used to identify true urgent cases. Several simulations were conducted to evaluate the performance of the proposed scheme. Results show that the suggested algorithm outperforms other ML based anomaly detection schemes presented in the literature and succeeds in addressing the BSN challenges.