Design, Implementation, and Evaluation of a Wearable Cardiorespiratory Fitness Monitor

Thesis Abstract

This study addresses the persistent gap in continuous, unobtrusive assessment of cardiorespiratory fitness (CRF) in real-world settings and the need for reliable wearable solutions that balance accuracy with user comfort. The objective is to design, implement, and evaluate a integrated wearable CRF monitor capable of estimating VO2max and real-time heart-rate–adjusted ventilatory thresholds through multi-sensor fusion while ensuring wearability and data integrity in daily activity. Specific objectives include (1) developing a sensor suite comprising photoplethysmography (PPG), impedance cardiography (ICG), accelerometry, and thermistor-based skin temperature to estimate VO2max and ventilatory thresholds; (2) implementing signal processing and machine-learning algorithms to calibrate and fuse data across modalities; (3) validating the device against a reference metabolic cart in controlled lab protocols and a target population; and (4) assessing user acceptance, durability, and data reliability in free-living conditions over a four-week period. The study adopts a multiphase methodological design combining engineering prototyping with rigorous validation and user-centered evaluation. The population comprises healthy adults aged 18–45 years (N=120) stratified by sex and physical activity level. An initial engineering phase recruits N=40 participants for laboratory testing, protocolized treadmill bouts with incremental workloads to obtain gold-standard VO2max via indirect calorimetry, ventilatory thresholds, and heart-rate profiles. A subsequent longitudinal field phase enrolls N=80 participants to wear the prototype during routine daily activities for four weeks, collecting continuous sensor data and weekly self-reports of perceived exertion and comfort. Data collection instruments include a multi-sensor wrist/arm wearable, a portable indirect calorimetry reference system for validation sessions, validated questionnaires (System Usability Scale, Comfort Rating Scales), and a mobile app for activity logging. Analytical approaches combine quantitative and qualitative methods. In the lab validation phase, regression-based calibration models, including multiple linear regression and random forest regression, are employed to map sensor features to VO2max and ventilatory thresholds, with performance assessed by RMSE, MAE, Bland–Altman limits of agreement, and concordance correlation coefficients. Sensor-level ablation studies identify the contribution of each modality. In the field phase, mixed-effects modeling evaluates the stability of estimates across days and activity types, while time-series analyses examine the temporal coherence between estimated CRF metrics and user-reported exertion. A subset of data (n?25) undergoes cross-validation with a repeated-measures ANOVA to examine device-consistency across sessions. The study is grounded in the theoretical framework of embodied cognition and the Ecological Momentary Assessment (EMA) paradigm, complemented by the Cardiorespiratory Fitness theory to justify multi-parameter estimation. The design also incorporates principles from the Technology Acceptance Model to interpret user feedback. Expected findings include high concordance between the wearable-derived VO2max estimates and indirect calorimetry (intraclass correlation coefficient >0.80) in lab settings, with acceptable error margins (RMSE < 5 mL·kg?1·min?1) after calibration. Ventilatory thresholds are anticipated to be detectable with sensitivity and specificity above 80% in controlled protocols. In free-living conditions, the device is expected to maintain robust performance with minimal signal loss during typical daily activities and demonstrate strong test-retest reliability (ICC > 0.75) across weeks. Qualitative results are projected to reveal favorable usability and comfort, with identified design improvements (strap tension, heat dissipation, battery life). The study contributes to knowledge by delivering a validated, multimodal wearable CRF monitor with demonstrable accuracy and reliability in both lab and real-world settings, advancing the integration of physiological sensing with consumer-grade wearables. It provides a replicable calibration pipeline and an open-source data-processing framework for cross-population generalization, and it offers evidence on how multi-sensor fusion can improve CRF estimation beyond single-modality approaches. Practical implications include enhanced remote monitoring of athletic training, clinical risk stratification, and population health surveillance, while methodological contributions encompass rigorous validation protocols for wearable physiological instruments and a transparent reporting standard for multi-sensor CRF estimation. Recommendations emphasize extending validation to diverse populations (older adults, clinical cohorts), improving power efficiency for longer wear, and exploring adaptive personalization of calibration models to maintain accuracy across lifestyle contexts.