In-situ calibration of accelerometers in body-worn sensors using quiescent gravity

Abstract

As the cost, size and power required by sensor devices decrease, an increasing range of applications are possible. We focus on the application of tracking accelerometer data from body worn sensors over long durations in health monitoring applications. Body worn sensors must be compact for the convenience of the patient and low power to support extended operation, but these benefits can come at the cost of accuracy. To mitigate this loss of accuracy we first examine the ability of calibration to improve accuracy. Next we propose an in-situ method for calibrating the accelerometers using the quiescent acceleration due to gravity as a calibration signal. This in-situ calibration uses only the stored data from the duration the sensor is worn by the patient and does not require any extra procedures or measurements from the physical sensor devices. Compared to a manual three-axis calibration technique, the proposed calibration can be applied to data without the need for specific calibration procedures or even access to the original sensor and provides comparable or better accuracy.