Analysis and detection of motion artifact in photoplethysmographic data using higher order statistics

In this work, we present novel methods for detecting the presence of motion artifact in photoplethysmographic (PPG) measurements based on higher order statistical information present in the data. We analyze both clean and corrupt PPG data in the time and frequency domains. In the time domain, skew and kurtosis measures of the signal are used as distinguishing metrics between clean and motion- corrupted data. In the frequency domain, the presence of random components due to motion artifact is analyzed using a frequency domain kurtosis measure. Additionally, bispectral analysis of PPG data indicates the presence of strong quadratic phase coupling (QPC) and more specifically self coupling in the case of clean PPG data. Though quadratic phase coupling is found in data corrupted by motion artifact, the self coupling feature is absent. A Neyman-Pearson (NP) detection rule is formulated for each of the measures. Additionally, treating each of the measures as observations from independent sensors, the Varshney-Chair rule is used to fuse individual decisions to form a global system decision.