Smart Sensors and Deep Learning for Recognizing Rehabilitation

Exercises

Parumanchala Bhaskar, Siddi Anitha, Siddannagari Susmitha, Katthigalla Sushma Chandrika,

Netla Harshita and Thapeta Anuradha

Department of Computer Science and Engineering, Santhiram Engineering College, Nandyal, Andhra Pradesh, India

Keywords: Smart Sensors, Graph Neural Networks, Rehabilitation Monitoring, Exercise Evaluation, Motion Analysis,

Deep Learning, Patient Recovery, Remote Healthcare, Wearable Technology.

Abstract: Rehabilitation can be quite a sensitive subject, needing to monitor it closely, especially for those recovering

from drug addiction and individuals who are undergoing physical therapy. The importance of rehabilitation

in basic terms is that structured rehabilitation exercises are the most essential and play a significant role in

physical and mental wellbeing by recovering strength, mobility and stability. In this paper, we propose a

method to identify and assess rehabilitation exercises based on the usage of smart sensors and Graph Neural

Network (GNN). They capture both spatial and temporal relationship in the movement data, thus improving

the accuracy of classifying exercises. Step Two: Utilizing wearable smart sensors to collect information about

the patient's movements and physiological parameters, and a GNN model to analyse the raw data and provide

feedback for teaching patients about their health status. This leads to high-confidence tracking, few errors, no

need for manual marking, and also allows for remote patient monitoring, which improves the overall

efficiency of rehabilitation. This not only stays relevant to drug-addicted individuals in control of their lives,

but it also keeps them engaged and on track to do this effectively with excellent rehabilitation programs during

their recoveries and patients with any kind of therapy-requiring care.

1 INTRODUCTION

Rehabilitation is one of the most important necessary

aspects that individuals need after suffering from

injuries, disabilities, or addiction. These are guided

exercises, designed to restore mobility, strength and

coordination. Rehabilitation traditionally has to be

supervised in-person by physiotherapists, who make

sure patients are doing their exercises correctly. This

approach, however, comes with challenges, including

high costs, accessibility problems, and a lack of

continuous observation.

The rapid development of smart sensors and deep

learning has enabled automated rehabilitation

tracking. Wearable devices have the potential to

enable continuous assessment of body movements,

physiological data recording, and exercise pattern

detection. These systems are real time and enable

patients to correct their mistakes immediately

leading to increase in therapy adherence.

Those who break this a refreshing way to relax and

get rid of stress, for example they have physical

exercises, metadative and others to recover the mental

health and body of them. Smart sensors can monitor

movements, heart rate and stress levels, assisting

professionals in assessing a patient’s progress. Deep

learning models could analyse this data for indicators

of improvement or relapses. These technologies give

tailored feedback, getting the patients to continue

with their procedure. Furthermore, remote

monitoring enables healthcare professionals to

support patients remotely, making rehabilitation more

accessible and effective.

1.1 Graph Neural Networks (GNNs) in

Rehabilitation

Rehabilitation Monitoring with Graph Neural

Networks (GNNs) A state-of-the-art deep learning

approach the connected graph modeling body

movements helps in accurate motion analysis and

precise classification of exercises. GNNs improve the

efficacy of rehabilitation by analyzing spatial and

temporal relationships in movement data. 3)

Bhaskar, P., Anitha, S., Susmitha, S., Chandrika, K. S., Harshita, N. and Anuradha, T.

Smart Sensors and Deep Learning for Recognizing Rehabilitation Exercises.

DOI: 10.5220/0013878300004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

111-115

ISBN: 978-989-758-777-1

111

Personalized Rehabilitation: The system adjusts

exercises according to patient progress, meaning an

optimized treatment plan is proposed, based on the

report of individual patient needs. In addition, the

system acts as an enhanced feedback system that

detects movement errors and allows for instantaneous

correction, which optimises the efficacy of exercise

while minimizing the risk of improper performance.

GNNs, with their complex architecture, improve the

scalability of the system as it is able to analyse no. of

rehabilitation exercises without the need of extensive

retraining for different conditions. Finally, this

system encourages better patient engagement by

providing interactive support and motivation to

foster adherence to rehabilitation programs over time.

GNN-based rehabilitation monitoring Toward these

advancements, a more efficient, accessible, and data-

driven rehabilitation process is ensured.

1.2 IMUs (Inertial Measurement Units)

And IMUs (Inertial Measurement Units) little

sensors that keep collect data about body motion via

acceleration, rotation, and direction. In rehabilitation

they assist in monitoring exercises without

cumbersome kit. IMUs, in conjunction with Graph

Neural Networks (GNNs), can better analyse

movements as it is able to capture different aspects of

the human skeleton. GNN GNN Link body parts to

data to assess exercise quality, abnormal motions, and

stress reactions. In addiction recovery, this

technology aids in tracking behaviors such as

stretching and breathing exercises, predicting risk for

relapse and providing immediate feedback.

Additionally, this allows for remote monitoring,

making rehabilitation more accessible and effective.

1.3 Heart Rate and GSR Sensors

These measure your heartbeat speed or sweat (GSR

or Galvanic Skin Response sensors) and stress or

relaxation. They monitor physical effort, and

emotional state, in rehabilitation.

The sensor data is processed in an ingenious

manner leveraging Graph Neural Networks (GNNs).

Each of the sensors are nodes, and GNNs link them

up to identify patterns between heart rate, stress levels

and movement. This is useful to recognizing signs of

anxiety, fatigue or risks of relapsing in addiction

recovery. Besides, GNN-based systems ensure the

rehabilitation process much more precise and

effective via real-time feedback and remoting of

control.

1.4 EEG Sensors

The patient wears EEG (electroencephalography)

sensors that pick up on electrical signals coming

from the scalp 2. And they track focus, stress and

emotional states, making them valuable in

rehabilitation and addiction recovery.

GNNs use a smart way to process the EEG data.

Among these techniques, we can highlight: Graph

neural networks: this step involves the transformation

of the brain into a network of nodes (each region of

the brain), where GNNs interlink these nodes,

allowing to learn patterns (a common characteristic

between these regions) that connect to an activity in

the brain. This aids in identifying stress, cravings or

improvement during addiction treatment. Finally,

GNN-based EEG systems foster rehabilitation

through personalization and effectiveness, thanks to

real-time insights and remote control.

2 LITERATURE SURVEY

The document explores how smart sensors and deep

learning improve rehabilitation by enabling better

exercise tracking and evaluation. Traditional rehab

requires supervision from doctors or physiotherapists,

which can be costly and not always accessible. Many

patients perform exercises at home, but without

proper monitoring, they may do them incorrectly or

lose motivation, leading to slower recovery and less

effective rehabilitation.

Indeed, the Smart Sensor-based Rehabilitation

Exercise Recognition (SSRER) system tackles the

aforementioned problem by monitoring body

movements through the use of wearable sensors,

thereby utilizing deep learning to analyze exercises.

These sensors consist of IMUs, which track your

movement, heart rate sensors to check your heart

health, GSR sensors that identify the skin changes

associated with stress, and EEG sensors that record

your brain activity, including emotional states.

Convolutional Neural Networks (CNNs) is the core

of the system for movement recognition, whereas

Gaussian Mixture Models (GMMs) can make data

more accurate, and Dynamic Convolutional Neural

Networks (D-CNNs) can accommodate differences in

speed and posture. Such technologies can offer real-

time feedback, monitor patients remotely, and

measure rehabilitation accurately, allowing recovery

to be more accessible, effective and tailor-made for

patients including those stuck at home.

Smart Sensor Based Rehabilitation System

Challenges Data synchronization is a challenge since

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

112

patients do the exercises at varying paces, making it

hard to align. Stimulus data with noise may produce

errors in the recognition of movements, and

variations of the same exercise will reduce accuracy.

The system is computationally intensive, preventing

real-time inference on low-power devices. Sensor

placement is not trivial; if sensors are misaligned,

results will differ. There are also new concerns

around health data being collected, and their clinical

effect needs to be validated their real-world

effectiveness will only be established once deployed.

Fixing these issues will improve system reliability

and patient outcomes.

3 PROPOSED SYSTEM

Table 1: Advantages of the Proposed Gnn-Based Rehabilitation

System Compared to Existing Systems.

Feature

Existing

Systems

Proposed GNN-

Based System

Movement

Tracking

Uses basic

motion

sensors, may

lack accuracy

Uses IMUs with

GNNs for precise

movement

recognition

Exercise

Classification

Limited to

predefined

movements

GNNs analyse

spatial and

temporal

movement

patterns,

improving

recognition

Personalization

Generic

exercise

plans, less

adaptive

Adjusts

rehabilitation

plans based on

patient progress

Error Detection

Requires

manual

supervision

Detects incorrect

movements

automatically and

provides real-

time corrections

Feedback

System

Delayed or

requires

clinical visits

Instant feedback,

improving

exercise

effectiveness

Stress & Mental

State

Monitoring

Rarely

included

Uses EEG and

GSR sensors to

track stress,

fatigue, and

emotional state

Overall

Effectiveness

Requires

frequent

supervision,

less adaptive

More accurate,

adaptive, and

cost-effective,

reducing recovery

time

The rehabilitation system proposed in this paper uses

the cutting-edge model of Graph Neural Networks

(GNNs) and includes sensor | graph} of IMUs, Heart-

rate (HR), GSR, and EEG. IMUs (Inertial

Measurement Units) track body movements to keep

posture and motion algorithms calculation straight.

For instance, if heart rate sensors measure physical

effort, gsr sensors detect stress by measuring the

sweat used by a user. Electroencephalogram (EEG)

sensors measure brain wave activity to Help measure

focus and emotions.

GNNs connect sensor data, treating each sensor

as nodes and examining the links that bind them. This

allows for accurate tracking of movements,

personalized adjustments to rehabilitation exercises,

and instant feedback on performance. The system

identifies motion errors, predicts future risk of stress

or relapse, and provides real-time corrective feedback

to maximize the efficiency of movement.

Rehabilitation becomes more accessible, interactive

and efficient as patients receive continuous support

through remote monitoring and adaptive exercise

plans. This is a smart and scalable solution that

delivers better recovery outcomes for everything

from physical therapy to addiction recovery.

The advantages of our system when compare to

existing system shown in above table 1.

4 MODULES

To implement the proposed approach using smart

sensors and Graph Neural Networks (GNNs) for

rehabilitation exercise recognition and evaluation,

figure 1 shows the System Architecture. The system

can be divided into the following key modules:

Figure 1: System Architecture.

Smart Sensors and Deep Learning for Recognizing Rehabilitation Exercises

113

4.1 Data Acquisition Module

• Wearable Smart Sensors: Analyse motion

and physiological data (e.g., accelerometers,

gyroscopes, EMG sensors, heart rate

monitors).

• Data Preprocessing: It includes noise

filtering, signal normalising, and extracting

features from raw sensor data.

4.2 Feature Extraction and

Representation Module

• Spatial and Temporal Data Processing:

Analyse motion patterns and body

physiology trends.

• Graph Construction: Model the body

movements as a graph in which nodes

represent sensor positions (e.g., joints) and

edges capture the spatial relationship.

4.3 Graph Neural Network (GNN)

Model Module

• Graph-Based Learning Use GNNs (e.g.

Graph Convolutional, Graph Attention

Networks) to analyze movement.

• Exercise Classification: Supervised or semi-

supervised learning to identify and classify

rehabilitation exercises.

4.4 Real-Time Feedback and

Recommendation Module

• Performance Evaluation: Compare patient

movements with standard exercise patterns.

• Personalized Feedback System: Provide

corrective guidance and suggestions based

on deviations.

• Adaptive Recommendations: Adjust

rehabilitation plans based on patient

progress.

4.5 Remote Monitoring and Patient

Engagement Module

• Cloud-Based Data Storage: Securely store

patient data for longitudinal tracking.

• Telemedicine Integration: Enable remote

monitoring and interaction with therapists.

• Gamification & Motivation: Implement

interactive rehabilitation exercises to

enhance engagement.

Figure 2 show the Model Accuracy Comparison for

Rehabilitation Exercise Recognition Systems.

Expected Output:

Figure 2: Model Accuracy Comparison for Rehabilitation

Exercise Recognition Systems.

Figure 3: Error Detection Rate Comparison.

Figure 4: Stress Monitoring Accuracy Comparison.

Figure 3 and 4 shows the Error Detection Rate

Comparison and Stress Monitoring Accuracy

Comparison respectively.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

114

5 CONCLUSIONS

It details a reforming framework with the help of keen

indicators and Graph Neural Networks (GNNs) to

screen and investigate activities for the individuals

who are recovering from wounds or drug habit.

Wearable sensors track body movements, heart rate,

stress and brain activity. GNNs able to process this

data on the fly to generate immediate feedback and

customized workout routines. The system includes

data collection, analysis in big data analytics, and

remote monitoring, which improves accuracy and

reduces errors, making rehabilitation more effective

and accessible. It provides improved movement

tracking, stress monitoring, and automated feedback

in comparison to traditional methods. Despite issues

on the level of data matching and sensor accuracy, the

ultimate benefit of the system is that people recover

better due to improved organization and efficiency of

the rehabilitation process.

REFERENCES

Chaitanya, V. Lakshmi, and G. Vijaya Bhaskar. "A priori

vs Genetic algorithms for Identifying Frequent Item

Sets." International journal of Innovative Research

&Development 3.6 (2014): 249-254.

Chaitanya, V. Lakshmi. "Machine Learning Based

Predictive Model for Data Fusion Based Intruder Alert

System." journal of algebraic statistics 13.2 (2022):

2477-2483

Chaitanya, V. Lakshmi, et al. "Identification of traffic sign

boards and voice assistance system for driving.” AIP

Conference Proceedings. Vol. 3028. No. 1. AIP

Publishing, 2024

Devi, M. Sharmila, et al. "Machine Learning Based

Classification and Clustering Analysis of Efficiency of

Exercise Against Covid-19 Infection." Journal of

Algebraic Statistics 13.3 (2022): 112-117.

Devi, M. Sharmila, et al. "Extracting and Analyzing

Features in Natural Language Processing for Deep

Learning with English Language." Journal of Research

Publication and Reviews 4.4 (2023): 497-502.

Mahammad, Farooq Sunar, Karthik Balasubramanian, and

T. Sudhakar Babu. "Comprehensive research on video

imaging techniques." All Open Access, Bronze (2019).

Mahammad, Farooq Sunar, and Y Madhu Viswanatham.

"Performance analysis of data compression algorithms

for heterogeneous architecture through parallel

approach." The Journal of supercomputing 76.4 (2020):

2275-2288.

Mahammad, Farooq Sunar, et al. "Key distribution scheme

for preventing key reinstallation attack in wireless

networks." AIP Conference Proceedings. Vol. 3028.

No. 1. AIP Publishing, 2024.

Mandalapu, Sharmila Devi, et al. "Rainfall prediction using

machine learning." AIP Conference Proceedings. Vol.

3028. No. 1. AIP Publishing, 2024.

Mr. M. Amareswara Kumar, effective feature engineering

technique for heart Disease prediction with machine

learning” in International Journal of Engineering &

Science Research, Volume 14, Issue 2, April-2024 with

ISSN 2277-2685.

Paradesi Subba Rao,”Detecting malicious Twitter bots

using machine learning” AIP Conf. Proc. 3028, 020073

(2024),https://doi.org/10.1063/5.0212693

Paradesi Subba Rao,"Morphed Image Detection using

Structural Similarity Index Measure”M6 Volume 48

Issue 4 (December 2024)

,https://powertechjournaI.com

Parumanchala Bhaskar, et al. "Machine Learning Based

Predictive Model for Closed Loop Air Filtering

System." Journal of Algebraic Statistics 13.3 (2022):

416-423.

Parumanchala Bhaskar, et al. "Incorporating Deep Learning

Techniques to Estimate the Damage of Cars During the

Accidents" AIP Conference Proceedings. Vol. 3028.

No. 1. AIP Publishing, 2024.

Parumanchala Bhaskar, et al “Cloud Computing Network

in Remote Sensing-Based Climate Detection Using

Machine Learning Algorithms” remote sensing in earth

systems sciences(springer).

Suman, Jami Venkata, et al. "Leveraging natural language

processing in conversational AI agents to improve

healthcare security." Conversational Artificial

Intelligence (2024): 699-711.

Sunar, Mahammad Farooq, and V. Madhu Viswanatham.

"A fast approach to encrypt and decrypt of video

streams for secure channel transmission." World

Review of Science, Technology and Sustainable

Development 14.1 (2018): 11-28.

Smart Sensors and Deep Learning for Recognizing Rehabilitation Exercises

115