BPqu: Breath Phase Quantifier

Citation Author(s):
Severin
Bernhart
Salzburg Research FGmbH
Submitted by:
Severin Bernhart
Last updated:
Thu, 03/20/2025 - 06:22
DOI:
10.21227/qkf5-vk58
Data Format:
Research Article Link:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

Endurance running is a popular activity due to its accessibility. However, participation is sometimes prevented by individuals experiencing respiratory problems. Monitoring breathing with body area networks can tackle these issues by tracking respiration during exercise and providing immediate, guiding feedback. Common breathing guidance systems rely on observational data from past breath cycles and consequently inherit disruptively lagging guidance interventions if breathing pattern suddenly change. Therefore, this work introduces BPqu, a deep learning algorithm for real-time breath phase estimation on edge devices that enhances new breathing guidance opportunities in running by predicting the continuous breath phase with zero latency. BPqu predicts breathing within the next running step and detects hale change events, determines inhalation and exhalation phases, and quantifies the concurrent continuous breath phase with performance metrics of 93.1%, 85.6%, and 81.5%, respectively, with zero delay. BPqu is designed to be deployed in a textile wearable breathing and stride sensor and will be applied to enhance intuitive,  non-distracting breathing guidance in running, reducing respiratory distress and increasing running enjoyment, thereby promoting human well-being.

Instructions: 

A detailed dataset documentation will be uploaded including script files once the related article is published.

"250212_BPQU_RESAMPLED_development_surrogates.csv" contains the raw and acceleration data aggregated from all participant runs within one CSV file. The data is already labelled with breathing regularity labels (attached, in-range, off-range, anomalous).

"250227_BPQU_CV_RESAMPLED_PREQ_casts_time_series.pkl" contains the continuous surrogate signals for each casting condition (back, now and fore).

"250318_BPQU_CV_RESAMPLED_PREQ_HALES.pkl" contains all preprocessed hales after applying the hale change event detection.

"250318_TF_BPQU_Preprossed_Evaluated_PhaseSurrogate.csv" contains preprocessed surrogate phases.

"250318_TF_BPQU_Preprossed_Evaluated_Reconstruction.csv" contains preprocessed reconstructed quantified breath phases.

"250318_BPQU_CV_RESAMPLED_PREQ.pkl" contains all evaluation performance metrics.

Funding Agency: 
Austrian Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology
Grant Number: 
2021-0.641.557