Comparative evaluation of adaptive filters in motion artifact cancellation for pulse oximetry

Among non-invasive approaches, pulse oximeters are most commonly used medical equipment for measuring blood oxygen saturation level. However, their accuracy is severely subjected to motion artifact and environmental noise. In this paper, we aim to evaluate empirically the effectiveness of adaptive filters in motion artifact cancellation for finger pulse oximeters. Our experiments compared the Least Mean Square (LMS) adaptive filter and the Exponentially Weighted Least Square (EWLS) adaptive filter with the Minimum Correlation Discrete Saturation Transform (MCDST). The experimental results indicate that both adaptive filters can perform better than the MCDST, and the EWLS adaptive filter better than the LMS adaptive filter in motion noise reduction.