Electrodermal Activity for Measuring Cognitive and Emotional Stress Level

Area of Science:

• Biomedical Engineering
• Psychophysiology
• Machine Learning

Background:

• Stress contributes to mental health issues like anxiety and depression.
• Electrodermal activity (EDA) is a noninvasive method for detecting stress and emotion.
• EDA signals require deconvolution to extract tonic and phasic components (skin conductance level and response).

Purpose of the Study:

• To develop and evaluate a stress level classification method using electrodermal activity (EDA) deconvolution.
• To compare the effectiveness of Continuous Deconvolution Analysis (CDA) and convex optimization approach to electrodermal activity (cvxEDA) for stress detection.
• To assess the accuracy of machine learning models in classifying stress levels based on EDA features.

Main Methods:

• Collected EDA signals from 18 healthy subjects during a Stroop test with varying stress levels.
• Deconvoluted EDA signals using Continuous Deconvolution Analysis (CDA) and cvxEDA.
• Extracted four statistical features from the deconvoluted signals (sample average, standard deviation, first absolute difference, normalized first absolute difference).
• Classified stress levels (mild, moderate, severe) using an Extreme Learning Machine (ELM) with 50 hidden layers.

Main Results:

• cvxEDA provided a smoother visualization of the phasic component compared to CDA.
• Both CDA and cvxEDA successfully separated skin conductance response (SCR) from raw signals and identified small peaks.
• The ELM model achieved high accuracy: 95.56% for CDA and 94.45% for cvxEDA.
• The developed system demonstrated that EDA can classify stress levels with over 94% accuracy.

Conclusions:

• EDA deconvolution combined with ELM provides an accurate method for stress level classification.
• The cvxEDA method offers a more precise visualization of SCR compared to CDA.
• This noninvasive system can aid in monitoring mental health and preventing stress-related disorders like anxiety and depression.