Unsupervised Motion Artifact Detection in Wrist-Measured Electrodermal Activity Data

TL;DR

This paper compares unsupervised and supervised machine learning methods for detecting motion artifacts in wrist-based electrodermal activity data, showing that unsupervised methods perform competitively and are more robust to additional sensor data.

Contribution

It demonstrates that unsupervised algorithms can effectively detect motion artifacts in EDA data across lab and real-world settings, challenging the reliance on supervised methods.

Findings

Unsupervised algorithms perform comparably to supervised ones in artifact detection.

Adding accelerometer data slightly improves supervised detection but worsens unsupervised performance.

Unsupervised methods are effective in real-world, long-duration EDA data analysis.

Abstract

One of the main benefits of a wrist-worn computer is its ability to collect a variety of physiological data in a minimally intrusive manner. Among these data, electrodermal activity (EDA) is readily collected and provides a window into a person's emotional and sympathetic responses. EDA data collected using a wearable wristband are easily influenced by motion artifacts (MAs) that may significantly distort the data and degrade the quality of analyses performed on the data if not identified and removed. Prior work has demonstrated that MAs can be successfully detected using supervised machine learning algorithms on a small data set collected in a lab setting. In this paper, we demonstrate that unsupervised learning algorithms perform competitively with supervised algorithms for detecting MAs on EDA data collected in both a lab-based setting and a real-world setting comprising about 23 hours of data. We also find, somewhat surprisingly, that incorporating accelerometer data as well as EDA improves detection accuracy only slightly for supervised algorithms and significantly degrades the accuracy of unsupervised algorithms.