Contrastive Learning of Electrodermal Activity Representations for Stress Detection

Katie Matton, Robert Lewis, John Guttag, Rosalind Picard
Proceedings of the Conference on Health, Inference, and Learning, PMLR 209:410-426, 2023.

Abstract

Electrodermal activity (EDA) is a biosignal that contains valuable information for monitoring health conditions related to sympathetic nervous system activity. Analyzing ambulatory EDA data is challenging because EDA measurements tend to be noisy and sparsely labeled. To address this problem, we present the first study of contrastive learning that examines approaches that are tailored to the EDA signal. We present a novel set of data augmentations that are tailored to EDA, and use them to generate positive examples for unsupervised contrastive learning. We evaluate our proposed approach on the downstream task of stress detection. We find that it outperforms baselines when used both for fine-tuning and for transfer learning, especially in regimes of high label sparsity. We verify that our novel EDA-specific augmentations add considerable value beyond those considered in prior work through a set of ablation experiments.

Cite this Paper

BibTeX
@InProceedings{pmlr-v209-matton23a, title = {Contrastive Learning of Electrodermal Activity Representations for Stress Detection}, author = {Matton, Katie and Lewis, Robert and Guttag, John and Picard, Rosalind}, booktitle = {Proceedings of the Conference on Health, Inference, and Learning}, pages = {410--426}, year = {2023}, editor = {Mortazavi, Bobak J. and Sarker, Tasmie and Beam, Andrew and Ho, Joyce C.}, volume = {209}, series = {Proceedings of Machine Learning Research}, month = {22 Jun--24 Jun}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v209/matton23a/matton23a.pdf}, url = {https://proceedings.mlr.press/v209/matton23a.html}, abstract = {Electrodermal activity (EDA) is a biosignal that contains valuable information for monitoring health conditions related to sympathetic nervous system activity. Analyzing ambulatory EDA data is challenging because EDA measurements tend to be noisy and sparsely labeled. To address this problem, we present the first study of contrastive learning that examines approaches that are tailored to the EDA signal. We present a novel set of data augmentations that are tailored to EDA, and use them to generate positive examples for unsupervised contrastive learning. We evaluate our proposed approach on the downstream task of stress detection. We find that it outperforms baselines when used both for fine-tuning and for transfer learning, especially in regimes of high label sparsity. We verify that our novel EDA-specific augmentations add considerable value beyond those considered in prior work through a set of ablation experiments.} }
Endnote
%0 Conference Paper %T Contrastive Learning of Electrodermal Activity Representations for Stress Detection %A Katie Matton %A Robert Lewis %A John Guttag %A Rosalind Picard %B Proceedings of the Conference on Health, Inference, and Learning %C Proceedings of Machine Learning Research %D 2023 %E Bobak J. Mortazavi %E Tasmie Sarker %E Andrew Beam %E Joyce C. Ho %F pmlr-v209-matton23a %I PMLR %P 410--426 %U https://proceedings.mlr.press/v209/matton23a.html %V 209 %X Electrodermal activity (EDA) is a biosignal that contains valuable information for monitoring health conditions related to sympathetic nervous system activity. Analyzing ambulatory EDA data is challenging because EDA measurements tend to be noisy and sparsely labeled. To address this problem, we present the first study of contrastive learning that examines approaches that are tailored to the EDA signal. We present a novel set of data augmentations that are tailored to EDA, and use them to generate positive examples for unsupervised contrastive learning. We evaluate our proposed approach on the downstream task of stress detection. We find that it outperforms baselines when used both for fine-tuning and for transfer learning, especially in regimes of high label sparsity. We verify that our novel EDA-specific augmentations add considerable value beyond those considered in prior work through a set of ablation experiments.
APA
Matton, K., Lewis, R., Guttag, J. & Picard, R.. (2023). Contrastive Learning of Electrodermal Activity Representations for Stress Detection. Proceedings of the Conference on Health, Inference, and Learning, in Proceedings of Machine Learning Research 209:410-426 Available from https://proceedings.mlr.press/v209/matton23a.html.

Related Material