Use of spectral estimation methods for computation of SpO<inf>2</inf> from artifact reduced PPG signals

Abstract

Arterial blood oxygen saturation (SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> ) is effectively measured by the pulse oximeter. The common cause of pulse oximeter failure, in error- free SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> estimation, is motion artifact (MA) corruption in the detected PPG signals. For a reliable and a low failure rate SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> estimation, the pulse oximeters must be provided with a clean artifact-free PPG signals with clearly separable DC and AC parts from which the SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> is computed in time domain. In this paper, we present non-parametric spectral estimation methods for computing SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> . The PPG signals recorded with frequently encountered artifacts (bending, vertical and horizontal motions of finger) were used for validation of the proposed methods. Experimental results revealed that the non-parametric spectral estimation methods are as accurate as the time domain analysis and particularly the Blackman-Tukey based SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> estimation out performed other non-parametric methods. Further, the Daubechies wavelet based method efficiently reduced motion artifacts restoring all the morphological features of the PPG signals.