A Preliminary Study on Automatic Motion Artifact Detection in Electrodermal Activity Data Using Machine Learning.

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Machine Learning Applications

Background:

• Electrodermal activity (EDA) is a key indicator of sympathetic nervous system function, widely used in emotion, stress, and medical condition assessments.
• Ambulatory EDA monitoring via wearable devices is growing, but susceptible to noise and motion artifacts.
• Accurate analysis of EDA signals necessitates robust automated detection of these artifacts.

Purpose of the Study:

• To develop and validate machine learning-based algorithms for detecting motion artifacts in electrodermal activity (EDA) signals.
• To address the challenge of noise and motion artifacts in ambulatory EDA recordings.

Main Methods:

• Collected simultaneous EDA signals from both hands of ten subjects, inducing motion in only one hand.
• Developed a cross-correlation-based method for unbiased labeling of EDA data segments.
• Extracted statistical, spectral, and model-based features, followed by feature selection and machine learning model training using a leave-one-subject-out validation approach.

Main Results:

• The proposed machine learning model achieved a classification accuracy of 83.85% with a standard deviation of 4.91% for motion artifact detection.
• The developed algorithm demonstrated superior performance compared to a recently published standard method.

Conclusions:

• Machine learning-based algorithms can effectively detect motion artifacts in electrodermal activity (EDA) signals.
• The proposed method offers a significant improvement for analyzing noisy EDA data from wearable sensors.