Data Fault Detection in Medical Sensor Networks

摘要: Medical body sensors can be implanted or attached to the human monitor physiological parameters of patients all time. Inaccurate data due sensor faults incorrect placement on will seriously influence clinicians’ diagnosis, therefore detecting has been widely researched in recent years. Most typical approaches fault detection medical area ignore fact that indexes aren’t changing synchronously at same time, and values mixed with abnormal illness make it difficult determine true faults. Based these facts, we propose a Data Fault Detection mechanism networks (DFD-M). Its includes: (1) use dynamic-local outlier factor (D-LOF) algorithm identify outlying sensed vectors; (2) linear regression model based trapezoidal fuzzy numbers predict which readings vector are suspected faulty; (3) proposal novel judgment criterion state according prediction values. The simulation results demonstrate efficiency superiority DFD-M.