Medical Instrumentation
Mohammad Saeed Zare Dehabadi; Mehran Jahed
Volume 10, Issue 3 , October 2016, , Pages 231-244
Abstract
Wireless Body Area Networks (WBAN) consist of a collection of biosensors utilized to remotely monitor the health status of patients. High accuracy anomaly detection and distinguishing between faults and physiological anomalies play a key role in proper detection of real emergency situations and is cruicial ...
Read More
Wireless Body Area Networks (WBAN) consist of a collection of biosensors utilized to remotely monitor the health status of patients. High accuracy anomaly detection and distinguishing between faults and physiological anomalies play a key role in proper detection of real emergency situations and is cruicial in lowering False Alarm Rate (FAR) cases. In this work, a univariate, unsupervised and real-time anomaly detection algorithm is proposed based on Hampel identifier and its performance is compared with previous and reported methods. Furthermore, a novel prediction method is introduced and utilized in order to correct for transient faults that are quite probable in WBANs, due to inherent noise and artifact of physiological sensors. Proposed method is shown to be faster than reported approaches while providing comparable. Final validation of the proposed method is performed by a real experimental dataset along with intentionally added faults and physiological anomalies. The results illustrate appropriate anomaly detection ability of the proposed approach.