ABSTRACT
This paper describes a wearable, open-source wrist temperature monitoring system that enables the reliable identification of slowly-varying skin temperature patterns that may be indicative of infections. The hardware platform uses a Bluetooth Low Energy (BLE) wireless interface and includes three skin temperature sensors, a thermally-isolated ambient temperature sensor, an inertial measurement unit (IMU), and a Galvanic skin response (GSR) sensor. A template-matching algorithm is used to detect weak but long-lived anomalous temperature patterns that deviate from the normal circadian rhythm are thus may be driven by infections. Experimental and simulation results confirm that small temperature anomalies (peak value <0.4°C) extending over 2-3 weeks can be detected with a total error rate <10%. IEEE