Thermal Imaging-based Breathing Phase Classification using Video Transformer

Fischerauer S, Dombrowski MN, Oesten M, Eskofier B, Richer R (2025)

Publication Status: Submitted

Publication Type: Unpublished / Preprint

Future Publication Type: Journal article

Publication year: 2025

DOI: 10.36227/techrxiv.176127403.32345542/v1

Abstract

Controlled breathing techniques are widely used in biofeedback and mental health interventions, but their effectiveness critically depends on reliable monitoring of breathing quality and adherence. Contactless camerabased systems offer a promising solution, yet most existing approaches remain limited to estimating average respiratory rate rather than capturing the full dynamics of the breathing process. In this work, we present the first end-to-end approach that reconstructs detailed respiratory signals and classifies all four phases of the breathing cycle – inhalation, post-inhalation pause, exhalation, and postexhalation pause – from thermal imaging. We introduce a new dataset with diverse breathing patterns under varying head movement conditions (HOLD and MOVE) and benchmark multiple state-of-the-art models adapted from related domains. Along this, we highlight previously underexplored challenges in thermal imaging: the ambiguity of breathing pauses due to low thermal contrast and the relationship between thermal signal, airflow and relative tidal volume. Our best-performing model, a video transformer, achieves Pearson correlations of 0.88 (HOLD) and 0.73 (MOVE) with respiration belt ground truth and macro F1 scores of 0.77 (HOLD) and 0.60 (MOVE) for four-phase breathing classification. This work helps make breathing exercises more measurable, interpretable, and ultimately more effective in real-world biofeedback applications.

Authors with CRIS profile

Sophie Fischerauer Machine Learning and Data Analytics Lab Mischa Neil Dombrowski Professur für Image Data Exploration and Analysis Marie Oesten Machine Learning and Data Analytics Lab Björn Eskofier Department Artificial Intelligence in Biomedical Engineering (AIBE) Robert Richer Machine Learning and Data Analytics Lab

How to cite

APA:

Fischerauer, S., Dombrowski, M.N., Oesten, M., Eskofier, B., & Richer, R. (2026). Thermal Imaging-based Breathing Phase Classification using Video Transformer. (Unpublished, Submitted).

MLA:

Fischerauer, Sophie, et al. Thermal Imaging-based Breathing Phase Classification using Video Transformer. Unpublished, Submitted. 2026.

BibTeX: Download