PULSE: Privileged Knowledge Transfer from Electrodermal Activity to Low-Cost Sensors for Stress Monitoring

2025-10-28

AI Summary

This work addresses the challenge of deploying costly electrodermal activity (EDA) signals for stress monitoring. We propose an EDA-guided cross-modal self-supervised pretraining framework that leverages EDA only during pretraining to learn modality-invariant representations. A shared-private embedding disentanglement mechanism enables effective knowledge transfer, while embedding alignment, multimodal fusion, and teacher–student knowledge distillation jointly ensure robust stress recognition using only low-cost sensors—ECG, BVP, ACC, and TEMP—at inference time. Evaluated on the WESAD dataset, our method significantly outperforms purely multimodal baselines, demonstrating that EDA-derived representations can be efficiently transferred to low-cost modalities. It achieves high accuracy (F1 > 0.89) while substantially reducing hardware dependency and deployment cost.

Technology Category

• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Multi-modal Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Electrodermal activity (EDA), the primary signal for stress detection, requires costly hardware often unavailable in real-world wearables. In this paper, we propose PULSE, a framework that utilizes EDA exclusively during self-supervised pretraining, while enabling inference without EDA but with more readily available modalities such as ECG, BVP, ACC, and TEMP. Our approach separates encoder outputs into shared and private embeddings. We align shared embeddings across modalities and fuse them into a modality-invariant representation. The private embeddings carry modality-specific information to support the reconstruction objective. Pretraining is followed by knowledge transfer where a frozen EDA teacher transfers sympathetic-arousal representations into student encoders. On WESAD, our method achieves strong stress-detection performance, showing that representations of privileged EDA can be transferred to low-cost sensors to improve accuracy while reducing hardware cost.

Problem

• Transfers stress detection from costly EDA sensors to affordable alternatives
• Enables inference using ECG, BVP, ACC, TEMP instead of EDA sensors
• Improves stress monitoring accuracy while reducing hardware costs

Innovation

• Self-supervised pretraining using EDA for initialization
• Separates encoder outputs into shared and private embeddings
• Transfers knowledge from EDA teacher to student encoders