A Signal Processing Technique to Reduce Artifact for NIRS

Abstract

Near infrared spectroscopy has fewer restrictions and lower electric noises than other brain function measurement methods such as EEG or MEG. Therefore, a measurement is easy and it is possible to measure the reaction of various kinds of tasks. However, the influence of changes in the cerebral blood flow often appears as an artifact on NIRS. In addition, the contribution ratio in each case of artifacts is unidentified. In this paper, we propose a technique to reduce or to eliminate the artifact by using the difference between the observed signal and artifact which is calculated by polynomial approximation. We tried to apply a polynomial approximation from the fourth to the sixth. As a result of the experiment, we confirmed that the artifact was reduced by our proposed technique, because the ripple content of differential signal closed to that of background signal.