| MS Seminar

Name of the Speaker: Mr. Sricharan Vijayarangan (EE21S068)
Guide: Dr. Mohanasankar S
Online meeting link: https://meet.google.com/oid-mwfb-tni
Date/Time: 9th May 2025(Friday), 3.00 PM
Title: A Machine Learning Driven Framework for Enhancing Cardiorespiratory Fitness Using Wearables.

Abstract :

This talk presents a comprehensive framework for precision cardiorespiratory monitoring and training through the extraction of physiological metrics from wearable ECG and accelerometer data. At its core lies a robust method for R-R interval estimation, a critical prerequisite for ECG-based analyses such as heart rate variability (HRV). Traditional R-peak detection methods often falter in ambulatory conditions due to motion artifacts and low signal quality. To address this, this work introduces a deep learning based regression approach that utilizes a distance map representation, allowing each ECG sample to encode its proximity to the nearest R-peak. This enables accurate valley-based localization of R-peaks even in noisy environments. The method was validated on public datasets and applied to real-world data from the Movesense wearable device, demonstrating strong generalizability beyond clinical contexts.

Building on accurate signal acquisition, we propose a novel framework for optimizing cardiorespiratory fitness (CRF), structured around three pillars: (1) Training Readiness (TR) Assessment, (2) Intelligent Dynamic Guidance Training, and (3) CRF Progress Measurement. TR is evaluated through the temporal instability of HRV, and a methodology is introduced to quantify and reduce this instability, thereby enhancing the predictive power of readiness assessments. To support effective training, an adaptive feedback system is developed that minimizes latency and maximizes adherence to prescribed protocols by dynamically adjusting exercise intensity based on real-time physiological data. For measuring CRF improvements, we evaluate the Cardiopulmonary Spot Jog Test (CPSJT) as a closed-loop, field-deployable proxy for VO2max. The CPSJT is benchmarked against established indicators such as resting heart rate and training consistency, demonstrating its utility for longitudinal progress tracking.

Together, these contributions form an end-to-end pipeline that transforms raw biosignal data into actionable feedback, enabling personalized, real-time interventions for sustainable improvements in cardiovascular fitness.