ε-Tube regression: a new method for motion artifact reduction

Abstract

This paper introduces a new regression method, called ε-tube regression (ε-TR), for motion artifact reduction in physiological signals. It forms a tube arround the data which leads to an approximation that models only the motion artifact and not the target signal. Moreover, ε-TR prescribes the shape of the approximation using the available information about the motion artifact. The results show that ε-TR can effectively remove the motion artifacts from the impedance signal measured on the arms.