Active Motion Artifact Reduction for Wearable Sensors Using Laguerre Expansion and Signal Separation

Abstract

This paper presents an active noise cancellation technique that utilizes an accelerometry measurement to recover corrupted wearable health monitoring signals. The technique presented here requires no calibration for each patient and is computationally efficient. Also, a method of determining when the desired signal is correlated with the motion reference is presented with a means of partial signal recovery. The Laguerre basis function provides robustness against calibration as well as improvement in computational efficiency, and symmetric decorrelation accommodates the management of correlated signals. Experimentation shows that the system can produce up to an 85% reduction in wearable sensor error for accelerations up to 2G.