Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Reinforcement Learning Empowered Activity-Aware Dynamic Health Monitoring Systems

Ziqiaing Ye, Yulan Gao, Yue Xiao, Zehui Xiong, Dusit Niyato

TL;DR

This work tackles energy-efficient, real-time health monitoring in smart environments by shifting from indiscriminate metric collection to activity-driven selective monitoring. It introduces DActAHM, a framework that fuses SlowFast-based activity recognition with a Deep Deterministic Policy Gradient policy to adapt health-metric monitoring to each user’s current activity, formalizing an optimization that jointly maximizes relevance and minimizes cost via a reward $\max\sum_n\big(R_n(a_n,F(a_n)) - \lambda C_n[r]\big)$. Key contributions include the formal problem formulation with cosine-similarity-based relevance, a DRL-based adaptive monitoring mechanism, and empirical results showing a 27.3% gain over the best fixed-action baseline and strong convergence behavior. The approach promises practical impact by reducing data processing and energy consumption in wearable- and vision-enabled health monitoring within urban settings while maintaining QoS. $

Abstract

In smart healthcare, health monitoring utilizes diverse tools and technologies to analyze patients' real-time biosignal data, enabling immediate actions and interventions. Existing monitoring approaches were designed on the premise that medical devices track several health metrics concurrently, tailored to their designated functional scope. This means that they report all relevant health values within that scope, which can result in excess resource use and the gathering of extraneous data due to monitoring irrelevant health metrics. In this context, we propose Dynamic Activity-Aware Health Monitoring strategy (DActAHM) for striking a balance between optimal monitoring performance and cost efficiency, a novel framework based on Deep Reinforcement Learning (DRL) and SlowFast Model to ensure precise monitoring based on users' activities. Specifically, with the SlowFast Model, DActAHM efficiently identifies individual activities and captures these results for enhanced processing. Subsequently, DActAHM refines health metric monitoring in response to the identified activity by incorporating a DRL framework. Extensive experiments comparing DActAHM against three state-of-the-art approaches demonstrate it achieves 27.3% higher gain than the best-performing baseline that fixes monitoring actions over timeline.

Deep Reinforcement Learning Empowered Activity-Aware Dynamic Health Monitoring Systems

TL;DR

This work tackles energy-efficient, real-time health monitoring in smart environments by shifting from indiscriminate metric collection to activity-driven selective monitoring. It introduces DActAHM, a framework that fuses SlowFast-based activity recognition with a Deep Deterministic Policy Gradient policy to adapt health-metric monitoring to each user’s current activity, formalizing an optimization that jointly maximizes relevance and minimizes cost via a reward . Key contributions include the formal problem formulation with cosine-similarity-based relevance, a DRL-based adaptive monitoring mechanism, and empirical results showing a 27.3% gain over the best fixed-action baseline and strong convergence behavior. The approach promises practical impact by reducing data processing and energy consumption in wearable- and vision-enabled health monitoring within urban settings while maintaining QoS. $

Abstract

In smart healthcare, health monitoring utilizes diverse tools and technologies to analyze patients' real-time biosignal data, enabling immediate actions and interventions. Existing monitoring approaches were designed on the premise that medical devices track several health metrics concurrently, tailored to their designated functional scope. This means that they report all relevant health values within that scope, which can result in excess resource use and the gathering of extraneous data due to monitoring irrelevant health metrics. In this context, we propose Dynamic Activity-Aware Health Monitoring strategy (DActAHM) for striking a balance between optimal monitoring performance and cost efficiency, a novel framework based on Deep Reinforcement Learning (DRL) and SlowFast Model to ensure precise monitoring based on users' activities. Specifically, with the SlowFast Model, DActAHM efficiently identifies individual activities and captures these results for enhanced processing. Subsequently, DActAHM refines health metric monitoring in response to the identified activity by incorporating a DRL framework. Extensive experiments comparing DActAHM against three state-of-the-art approaches demonstrate it achieves 27.3% higher gain than the best-performing baseline that fixes monitoring actions over timeline.
Paper Structure (12 sections, 6 equations, 4 figures, 1 algorithm)

This paper contains 12 sections, 6 equations, 4 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. System Overview
  4. Dynamic Monitoring Strategy Model
  5. Relevance and Cost of Monitoring Metrics
  6. Problem Formulation
  7. DActAHM Design
  8. State Space
  9. Action Space
  10. Reward
  11. Simulation Result
  12. Conclusion

Figures (4)

  • Figure 1: A comprehensive overview of a sophisticated smart healthcare system operating in an urban setting.
  • Figure 2: DDPG algorithm convergence verification.
  • Figure 3: Algorithm comparison in static scenario.
  • Figure 4: Algorithm comparison in dynamic scenario.