Title:

“ SMART HEALTH GUARD: AI-POWERED VOICE-

ACTIVATED EMERGENCY RESPONSE AND HEALTH
MONITORING APP”

A CORE COURSE PROJECT REPORT

Submitted By

AKSHAYA R REG NO. 213223104008

GAYATHRI S REG NO. 213223104047

in partial fulfilment for the award of the degree of

BACHELOR OF ENGINEERING
IN

COMPUTER SCIENCE AND ENGINEERING

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

CHENNAI INSTITUTE OF TECHNOLOGY AND APPLIED
RESEARCH
Sarathy Nagar, Kundrathur, Chennai-600069
OCT / NOV – 2025
Vision of the Institute:

To be an eminent centre for Academia, Industry and Research by imparting

knowledge, relevant practices and inculcating human values to address global
challenges through novelty and sustainability.

Mission of the Institute:

 DEPARTMENT OF
COMPUTER SCIENCE AND ENGINEERING

Vision of the Department:

To Excel in the emerging areas of Computer Science and Engineering by imparting

knowledge, relevant practices and inculcating human values to transform the students
as potential resources to contribute innovatively through advanced computing in real
time situations.

Mission of the Department:

DM1. To provide strong fundamentals and technical skills for Computer Science
applications through effective teaching learning methodologies.
DM2. To transform lives of the students by nurturing ethical values, creativity and
novelty to become Entrepreneurs and establish start-ups.
DM3. To habituate the students to focus on sustainable solutions to improve the quality
of life and the welfare of the society.
 CERTIFICATE

This is to certify that the “Core Course Project” Submitted by AKSHAYA R

(Reg no: 213223104008) and GAYATHRI S (Reg no: 213223104047) is a work done
by him/her and submitted during 2025-2026 academic year, in partial fulfilment of the
requirements for the award of the degree of BACHELOR OF ENGINEERING in
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING, at Chennai
Institute of Technology and Applied Research.

Dr.S.Muthumarilakshmi
Project Coordinator Internal Examiner

Dr.S.Pavithra
Head of the Department External Examiner
 ACKNOWLEDGEMENT

We express our gratitude to our Chairman Shri.P.SRIRAM and all trust members of
Chennai institute of technology for providing the facility and opportunity to do this
project as a part of our undergraduate course.

We are grateful to our Principal Dr.A.RAMESH M.E, Ph.D. for providing us the
facility and encouragement during the course of our work.

We sincerely thank our Head of the Department Dr.A.Pavithra, Department of

Computer Science and Engineering for having provided us valuable guidance,
resources and timely suggestions throughout our work.

We would like to extend our thanks to our Project Co-ordinator Dr. Muthumari
Lakshmi, Department of Computer Science and Engineering, for her valuable
suggestions throughout this project.

We wish to extend our sincere thanks to all Faculty members of the Department of
Computer Science and Engineering for their valuable suggestions and their kind
cooperation for the successful completion of our project.

We wish to acknowledge the help received from the Lab Instructors of the Department
of Computer Science and Engineering and others for providing valuable suggestions
and for the successful completion of the project.

NAME: AKSHAYA R REG.NO: 213223104008

NAME: GAYATHRI S REG.NO: 213223104047
 PREFACE

I, a student in the Department of Computer Science and Engineering, need to

undertake a project to expand my knowledge. The main goal of my Core Course Project
is to acquaint me with the practical application of the theoretical concepts I’ve learned
during my course.

It was a valuable opportunity to closely compare theoretical concepts with real-

world applications. This report may depict deficiencies on my part but still it is an
account of my effort.

The results of my analysis are presented in the form of an industrial Project, and
the report provides a detailed account of the sequence of these findings. This report is
my Core Course Project, developed as part of my 3rd year project. As an engineer, it is
my responsibility to contribute to society by applying my knowledge to create
innovative solutions that address their changes.
 DECLARATION

I hereby declare that this project titled “Smart Health Guard: AI-
Powered Voice-Activated Emergency Response and Health Monitoring
App”,is our original work. It has been completed in accordance with the
guidelines provided by Chennai Institute of Technology and Applied
Research. This project has not been submitted for any other degree or
diploma, and all sources and references used in the preparation of this
project have been acknowledged appropriately. I affirm that the ideas and
expressions herein are our own and do not infringe upon the rights of any
other author or researcher.

This project represents our independent research and analysis. I confirm

that the findings, conclusions, and recommendations contained within this
document are based on our own work and insights. We have conducted
thorough research and adhered to the highest standards of academic
integrity throughout the process. This work is original and has not been
previously published or submitted elsewhere. We take full responsibility for
the content and quality of this project.
 ABSTRACT

The increasing need for reliable and accessible healthcare solutions has
emphasized the importance of real-time monitoring and rapid emergency
response, especially for elderly individuals and those with chronic
conditions. Traditional healthcare support often suffers from delayed
detection, lack of integration, and limited accessibility. To address these
challenges, this project presents Smart Health Guard, an AI-powered,
voice-activated application for health monitoring and emergency assistance.

The platform integrates fall detection, heart rate monitoring using

Photoplethysmography (PPG), medication reminders, AI chatbot
support, and smart appliance control to provide a comprehensive
healthcare solution. In emergencies, the system automatically triggers loud
alarms, live location sharing, and emergency calls, ensuring timely help
without intermediaries.

By combining AI, IoT, and voice recognition technologies, Smart Health

Guard enhances safety, accessibility, and independence for vulnerable
populations. This project demonstrates how technology-driven solutions
can improve healthcare delivery through proactive monitoring, rapid
response, and secure, user-friendly design.
 1.INTRODUCTION

1.1 BACKGROUND OF THE STUDY

The continuous rise in life expectancy and the growing prevalence of

chronic diseases such as diabetes, hypertension, cardiovascular
disorders, and respiratory illnesses have placed a substantial burden on
global healthcare systems. According to the World Health Organization
(WHO), the population aged 60 years and above is expected to double by
2050, reaching over 2 billion worldwide. This demographic transition
presents significant healthcare challenges, particularly in terms of
accessibility, affordability, and continuous monitoring of vulnerable
populations.

Elderly individuals and patients with long-term medical conditions are

especially at risk of critical medical emergencies. Incidents such as falls,
strokes, cardiac arrests, and sudden fluctuations in vital signs often
require immediate medical intervention. However, delays in recognition
or response to such emergencies can result in severe complications,
irreversible health deterioration, or even death. Studies show that falls
are the second leading cause of accidental injury deaths worldwide,
with older adults being most affected. Similarly, conditions like
hypertension and diabetes often progress silently until a sudden crisis
occurs, making early detection and intervention essential.
Despite ongoing improvements, traditional healthcare systems face
several limitations. In many cases, monitoring is episodic rather than
continuous, with patients only receiving medical attention during
scheduled check-ups or after a visible crisis. Emergency detection often
relies on manual reporting by patients or wearable devices, which may
not be affordable, comfortable, or user-friendly for elderly users.
Furthermore, caregivers and healthcare professionals cannot be physically
present at all times, leading to a critical gap in continuous healthcare
support. This lack of timely monitoring and response increases the
vulnerability of elderly individuals and places additional stress on
caregivers and healthcare systems.

In recent years, advances in Artificial Intelligence (AI), Internet of

Things (IoT), and mobile health applications (mHealth) have created
new opportunities for personalized, accessible, and proactive healthcare
solutions. These technologies make it possible to analyse health data in real
time, recognize abnormal patterns, and trigger instant alerts or automated
interventions. For example, AI-based models can detect unusual heart
rhythms or sudden movement patterns associated with falls, while IoT-
enabled devices can connect with emergency services or caregivers within
seconds. Mobile health applications further enhance accessibility by
bringing these technologies directly to smartphones, which are widely
available and cost-effective.
Building on these innovations, Smart Health Guard has been designed as a
comprehensive AI-powered mobile application that integrates fall
detection, real-time monitoring of vital signs, abnormal health alerts,
medication reminders, AI-driven chatbot support, and smart home
integration. Unlike many traditional solutions, it minimizes dependency on
external hardware by utilizing the smartphone’s built-in sensors
(accelerometer, gyroscope, and camera).

Its focus on ease of use, affordability, and accessibility ensures that

healthcare support can reach a wide range of users, particularly elderly
individuals, patients with chronic illnesses, and those living alone.
By combining AI intelligence, IoT connectivity, and mobile
accessibility, Smart Health Guard addresses the limitations of traditional
healthcare and creates a holistic, proactive healthcare ecosystem. It aims to
not only improve emergency response times but also enhance preventive
care, independence, and trust between patients, families, and healthcare
providers.

1.2 RESEARCH PROBLEM

While there has been remarkable progress in healthcare technology, critical
gaps still exist in ensuring instant emergency detection, continuous
monitoring, and accessibility for vulnerable populations. Existing systems
often fail to provide an integrated, affordable, and user-friendly solution.
The main problems identified are:
 Dependence on external devices – Most health monitoring systems
rely on wearables (e.g., smartwatches, fitness bands) to track vitals or
detect emergencies. These devices, however, are expensive, require
charging, and are often ignored or misused by elderly individuals.
This dependency creates barriers to adoption.
 Lack of integration – Current solutions usually address one specific
feature such as heart rate monitoring or fall detection. Few provide a
comprehensive healthcare ecosystem that integrates multiple
features like health monitoring, emergency alerts, medication
reminders, and smart home support within a single platform.
 Delayed emergency response – In many situations, users or
caregivers must manually trigger alerts when emergencies occur.
During sudden episodes such as fainting, stroke, or cardiac arrest, this
manual action is not possible, leading to critical delays that may
result in loss of life.
 Accessibility challenges – Many healthcare applications are not
designed with elderly users in mind. They often include small text
sizes, complicated navigation, dependence on stable internet
connectivity, or limited local language support. These usability issues
reduce the effectiveness of existing systems in real-world scenarios.
 Data privacy and security concerns – Many healthcare applications
rely on cloud storage or third-party platforms for handling sensitive
health data. This creates concerns about data breaches,
unauthorized access, and loss of patient confidentiality. Elderly
 users and patients may hesitate to adopt such solutions due to a lack
of trust in how their personal health information is managed.
Therefore, there is a need for a healthcare system that ensures local
data storage, strong encryption mechanisms, and user-controlled
sharing of information to balance accessibility with security.

1.2 RESEARCH QUESTIONS / OBJECTIVES

RESEARCH QUESTIONS:
1. How can AI and IoT technologies be combined to create a
healthcare platform that ensures real-time monitoring and rapid
emergency response?
2. Can a mobile application accurately detect emergencies such as
falls, irregular heartbeats, or abnormal vitals without relying on
external hardware?
3. How can voice-activated features improve the usability and
accessibility of healthcare applications for elderly individuals?
4. What is the potential impact of such a system on improving safety,
independence, and caregiver support?

The OBJECTIVES of this study are to:

1. Design and develop a mobile application that integrates real-time
health monitoring and emergency detection.
2. Implement AI-based fall detection using smartphone sensors such
as the accelerometer and gyroscope.
 3. Measure heart rate through the smartphone’s camera and
flashlight using the Photoplethysmography (PPG) technique,
eliminating the need for external devices.
4. Incorporate voice recognition, IoT-based smart appliance control,
and an AI health chatbot to create a comprehensive healthcare
ecosystem.
5. Provide medication reminders with interactive notifications to
encourage regular treatment adherence.
6. Evaluate the system’s accuracy, usability, and efficiency through
prototype testing with both elderly and general users.

1.4 SIGNIFICANCE OF THE STUDY

This research is significant because it addresses one of the most pressing
challenges in modern society: ensuring accessible, affordable, and
reliable healthcare for an aging global population. Its contributions are
relevant to multiple stakeholders:
 For elderly individuals – The system offers independence, safety,
and confidence, ensuring that help is always available during
emergencies without requiring constant caregiver presence.
 For patients with chronic conditions – Continuous monitoring
reduces the risk of undetected deterioration and helps ensure better
treatment compliance.
 For caregivers and family members – Automated alerts, live
location sharing, and emergency notifications reduce the burden of
 continuous monitoring, providing peace of mind and enabling
remote caregiving.
 For healthcare systems – By encouraging preventive care and
timely interventions, the solution can reduce hospital admissions,
lower treatment costs, and ease the burden on overworked healthcare
providers.
By integrating AI-driven detection, instant communication, and smart
home connectivity, Smart Health Guard demonstrates its potential as a
scalable, affordable, and practical solution that can be deployed globally
across diverse healthcare settings.

1.5 SCOPE OF THE STUDY

The scope of this project is to design, develop, and test a mobile-based
healthcare application that uses AI, IoT, and voice recognition
technologies to provide real-time monitoring and emergency support.
Features within scope include:
 Fall detection using accelerometer and gyroscope data.
 Heart rate monitoring through camera + flashlight (PPG method).
 Abnormal vital sign alerts using AI-based analysis.
 An AI-powered health chatbot for answering common health
queries.
 Medication reminders with push notifications and tracking.
 IoT-enabled smart appliance control (e.g., fans, lights, AC) to
improve accessibility.
Limitations / exclusions from scope:
 The current prototype works primarily with smartphone-based
sensors. Full wearable integration (e.g., smartwatches, medical-
grade devices) is proposed for future work.
 Health data is stored locally to ensure privacy. Cloud-based
synchronization and predictive AI models for long-term health
forecasting will be part of future expansions.
 Testing is currently limited to prototype evaluation with a small user
base. Large-scale clinical trials and regulatory approvals are outside
the scope of this research.
 2.LITERATURE REVIEW

2.1 REVIEW OF RELEVANT PREVIOUS WORK

Healthcare technologies have evolved rapidly with the integration of
Artificial Intelligence (AI), Internet of Things (IoT), and mobile health
(mHealth) applications, offering promising solutions for remote
monitoring, emergency detection, and chronic disease management.

A review of the literature highlights several streams of research:

 AI in Healthcare Monitoring
Artificial Intelligence has been widely applied in health monitoring
for anomaly detection and predictive analysis. Machine learning
algorithms can detect irregularities in heart rhythms, oxygen levels,
and blood pressure, while Convolutional Neural Networks
(CNNs) are used in fall detection by analyzing patterns in
accelerometer and gyroscope data. For example, studies have
demonstrated that CNN-based models can achieve high accuracy in
distinguishing between normal movements and falls, enabling
automated detection and early intervention.

 IoT-enabled Healthcare Systems

IoT has transformed healthcare by connecting sensors, wearables,
and smartphones into a unified system. IoT devices can
continuously capture patient vitals and transmit them to caregivers or
cloud servers for real-time analysis. Applications include remote
 patient monitoring, telemedicine, and smart home health
integration. For instance, IoT-based glucose monitoring systems help
diabetic patients track blood sugar levels and share data instantly with
doctors. However, these systems are often limited by high device
cost, dependence on internet connectivity, and privacy concerns.

 Fall Detection Technologies

Falls are among the most dangerous health risks for elderly
individuals. Existing systems include wearable sensors, vision-based
systems (CCTV or cameras), and smartphone-based detection.
While wearables provide accurate results, they face issues of
compliance (users forget to wear them) and cost. Vision-based
systems raise privacy issues and are unsuitable in all environments.
Smartphone-based detection has emerged as a practical alternative, as
most users carry phones equipped with accelerometers and
gyroscopes. Studies show that fall detection accuracy improves when
combined with AI algorithms and threshold-based methods.

 Mobile Health (mHealth) Applications

The rapid growth of smartphone usage has made mHealth
applications one of the most accessible healthcare solutions. Apps
are now used for medication reminders, chronic disease tracking,
teleconsultations, and mental health support. For example, heart
rate can be measured using Photoplethysmography (PPG) through a
 smartphone camera and flashlight, eliminating the need for external
devices. While these apps are affordable and widely available, many
lack integration with emergency response features and elder-
friendly design.

2.2 THEORETICAL FOUNDATIONS

The Smart Health Guard system is grounded in several theoretical
frameworks:
1. Artificial Intelligence (AI) Theory
AI provides the theoretical basis for enabling machines to learn,
classify, and predict outcomes from datasets. In healthcare,
supervised learning models are applied to classify normal vs.
abnormal health patterns, while deep learning models such as CNNs
are effective in motion analysis for fall detection.
2. Internet of Things (IoT) Theory
IoT theory emphasizes the interconnection of physical devices
through networks for real-time data sharing. In healthcare, this
translates to continuous monitoring using sensors and mobile devices
that communicate with emergency services, caregivers, and smart
home systems.
3. Human-Computer Interaction (HCI)
HCI focuses on designing systems that are usable, accessible, and
user-centered. For elderly care, this means designing healthcare apps
with voice commands, simplified navigation, and large text
interfaces to minimize usability challenges.
4. Mobile Health (mHealth) Framework
The mHealth approach emphasizes the role of mobile devices in
healthcare delivery, especially in resource-limited settings. It
provides a cost-effective and scalable solution to extend healthcare
access to wider populations.
5. Telemedicine and Remote Care
Telemedicine theories provide the foundation for remote health
monitoring and consultations, allowing patients to receive
healthcare without being physically present in clinics. This is
particularly useful for rural and underserved populations.

2.3 GAPS IN THE LITERATURE

Despite progress, several gaps remain in existing healthcare
solutions:
 Device Dependency – Most health monitoring solutions depend on
external devices (e.g., smartwatches, sensors, or wearables). While
effective, these devices are costly, need regular maintenance, and are
not always adopted by elderly users.
 Feature Fragmentation – Many solutions focus only on one
problem, such as fall detection, heart rate monitoring, or
medication reminders. Few offer a holistic ecosystem that integrates
multiple features into one platform.
 Elder Accessibility – Usability challenges remain a major barrier.
Many apps are designed for younger users and lack elder-friendly
 features like voice commands, simplified navigation, and offline
capabilities.
 Privacy and Data Security – Sensitive health data is often stored in
cloud servers, creating risks of data breaches and raising concerns
about trust and confidentiality.
 Emergency Response Delays – Several systems depend on manual
alerts triggered by patients or caregivers. During sudden crises such
as fainting, this is not possible, leading to life-threatening delays.
These gaps underscore the importance of designing a mobile-based,
AI-powered, voice-activated system like SmartHealthGuard that
addresses integration, accessibility, and privacy concerns while
ensuring real-time emergency response.

2.4 HYPOTHESES OR RESEARCH FRAMEWORK

Based on the literature review and gaps, this study proposes the
following hypotheses:
 Hypothesis 1: A mobile-based AI-powered healthcare system can
provide accurate fall detection and vitals monitoring without external
devices.
 Hypothesis 2: Integration of voice recognition and AI chatbots will
significantly improve accessibility and user adoption among elderly
users.
 Hypothesis 3: Combining AI, IoT, and mHealth into a unified
system will result in faster and more reliable emergency response
compared to traditional methods.
 Hypothesis 4: Local data storage and user-controlled sharing will
increase trust and security, encouraging adoption of mobile health
solutions.
Research Framework
The framework for SmartHealthGuard combines:
1. AI-based health monitoring – fall detection using smartphone
sensors, heart rate detection using PPG.
2. IoT-enabled integration – smart appliances and emergency
connectivity (lights, fans, emergency numbers).
3. Voice-enabled accessibility – voice commands for emergency alerts
and smart home control.
4. Preventive care features – medication reminders, AI chatbot for
health queries.
5. Emergency management – automated alerts, alarms, live location
sharing, and emergency calls.
 3.METHODOLOGY

3.1 RESEARCH DESIGN (ARCHITECTURE / FRAMEWORK)

The architecture of Smart Health Guard is designed to function as a

real-time health monitoring and emergency response system. It
begins with a health monitoring device, primarily a smartphone
equipped with sensors such as an accelerometer, gyroscope, and
camera with flashlight. These sensors continuously capture vital
parameters and user activity.
 The data collected is then passed to the AI-based emergency
detection module, which evaluates whether the readings indicate
normal conditions or an abnormal/emergency situation. If no
emergency is detected, the system continues monitoring in the
background. If an emergency is detected — such as a fall, abnormal
heart rate, or other irregular vitals — the system immediately initiates
the alert mechanism.

The alert mechanism is designed to send notifications via voice

output, SMS, and in-app alerts, ensuring that caregivers or
emergency services are instantly informed. In addition to
communication alerts, the system integrates with IoT-enabled smart
appliances, allowing automated actions like switching on lights,
unlocking doors, or activating alarms. These features ensure that
emergency responders can access the environment quickly and safely.

Thus, the framework works in a closed loop:

 Input: Health data and activity data from sensors.
 Processing: AI-driven emergency detection.
 Output: Alerts + smart home activation.

3.2 DATA COLLECTION METHODS

The system collects data through a combination of real-time sensor
monitoring and user-defined inputs.
  Sensor Data:
The smartphone accelerometer and gyroscope provide continuous
motion data, which is analyzed for fall detection. The camera and
flashlight record fingertip blood flow using Photoplethysmography
(PPG) to estimate heart rate.
 Emergency Detection Data:
Data is collected whenever the AI module flags abnormal vitals or
movement patterns. This helps in evaluating the accuracy and
response time of the emergency detection process.
 User Input Data:
Users provide medication schedules, which are stored and used to
generate reminders. User interactions with the AI chatbot are also
logged for evaluating support effectiveness.
This combined dataset ensures that the system can be tested both in
normal conditions (no emergency) and critical conditions
(emergency triggered).

3.3 TOOLS, MATERIALS, AND PROCEDURES USED

 Tools and Technologies:

The application is built using Flutter for cross-platform development,
Syncfusion charts for data visualization, and APIs for
communication (HTTP/MQTT). AI-based fall detection is
implemented with a CNN model, while the camera + flashlight is
 used for PPG heart rate monitoring. For alerts, APIs enable voice
calls, SMS, and push notifications.

 Materials:
A smartphone with built-in sensors (accelerometer, gyroscope,
camera, and flashlight) is the only essential hardware. For extended
functionality, Wi-Fi or Bluetooth-enabled smart appliances are
used for IoT integration.

 Procedures:
o Sensor data is continuously collected.
o AI algorithms analyze the data for abnormalities.
o If detected, the system automatically triggers alerts and activates
IoT devices.
o If not, monitoring continues seamlessly in the background.

3.4 DATA ANALYSIS METHODS

The data analysis focuses on evaluating whether the system works
accurately and reliably.
 Sensor Data Analysis: The accelerometer and gyroscope readings are
analyzed to distinguish between normal movements and falls.
Abnormal readings trigger the CNN fall detection algorithm.
 Vitals Analysis: The PPG data is processed to calculate beats per
minute (BPM). This is compared against normal ranges to detect
irregular heart patterns.
  Alert Efficiency: Each emergency event is measured for response
time (time from detection to alert delivery). The system also tracks
whether the SMS, voice alert, or IoT activation was successful.
This analysis ensures that both the health monitoring and alert
systems are validated against real-world use cases.

3.5 ALGORITHM / PROCEDURE / PSEUDO CODE

The system follows a straightforward algorithm:
1. Fall Detection (AI-CNN model)
o Input: Accelerometer and gyroscope readings.
o Process: Normalize data → segment into time windows → run
CNN classification.
o Output: Fall detected / Not detected.

Step 1: Collect motion data (x, y, z axes)

Step 2: Segment into time windows
Step 3: Preprocess (normalize, filter noise)
Step 4: Feed into CNN model
Step 5: Classify activity as NORMAL / FALL
If FALL detected:
Trigger Emergency Alert ()
Else:
Continue Monitoring

2. Heart Rate Monitoring (PPG)

 o Input: Video frames captured using camera + flashlight.
o Process: Measure intensity changes → apply signal processing
→ calculate BPM.
o Output: Current heart rate with visualization.
Step 1: Illuminate fingertip using flashlight
Step 2: Record colour intensity changes caused by blood flow
Step 3: Apply Fast Fourier Transform (FFT) to extract periodic signals
Step 4: Calculate Beats Per Minute (BPM)
Step 5: Display BPM & Plot Graph
o

3. Emergency Response
If Emergency Detected:
Send Voice Alert ()
Send SMS + App Notification ()
Share Live Location ()
Activate IoT Devices ()
Else:
Continue Monitoring

o If emergency detected → Send alerts (Voice/SMS/App) →

Share live location → Activate smart appliances.
o If no emergency → Continue background monitoring.

3.6 ETHICAL CONSIDERATIONS

 As the project involves sensitive health data, ethical considerations
are essential:
 Data Privacy: Health data is stored locally on the user’s device. Only
with user permission can it be shared externally.
 Consent: Users must explicitly allow the app to access sensors,
camera, flashlight, and location services.
 False Alarms: To avoid panic, users are given the option to cancel
alerts with a confirmation button if the system misclassifies a
situation.
 Accessibility: The app design prioritizes elderly users by providing
voice commands, simple navigation, and large text.
 User Safety: IoT integration (like unlocking doors) is designed
carefully to avoid compromising home security.
 4.RESULTS / FINDINGS
4.1 PRESENTATION OF DATA / RESULTS
The Smart Health Guard application was successfully developed,
deployed on Android devices, and tested under multiple scenarios to
evaluate its functionality, reliability, and usability.

The findings are summarized across the core modules as follows:

Emergency Monitoring
 The emergency monitoring feature allows users to configure
emergency contact numbers and set predefined time intervals for
periodic status checks.
 Once monitoring begins, the system continuously observes the user’s
activity.
 If the app detects abnormal activity (such as sudden immobility or
potential falls) or extended inactivity within the set interval, the user
receives a prompt.
 The “I am Okay” button enables the user to quickly dismiss false
alarms. This prevents unnecessary notifications being sent during
non-critical events.
  In the absence of a response within a defined time frame, the system
automatically triggers multi-channel alerts.
 Alerts were tested in real-time and successfully reached emergency
contacts, proving the reliability of the monitoring cycle.

Heart Rate Monitoring

 Heart rate detection was implemented using Photoplethysmography
(PPG) with the smartphone’s camera and flashlight.
 During testing, users placed their fingertip over the camera, allowing
the system to measure blood flow intensity variations.
 The module provided the following outputs:
o Maximum Heart Rate recorded during the session.
o Minimum Heart Rate recorded.
o Average Heart Rate over the monitoring period.
 Results were displayed in both:
o Numerical values (e.g., Max: 96 BPM, Min: 90 BPM, Avg: 93
BPM).
o Graphical line charts showing Heart Rate Over Time,
enabling users to track trends and detect fluctuations visually.
 Historical data was logged and presented in a measurement history
section, which allowed users to review their past readings and verify
improvements or abnormalities over time.
 Testing confirmed that BPM readings were consistent and had only
minor deviations compared to medical-grade pulse oximeters.
Alert System
 The emergency alert system was tested with different simulated
conditions (fall, abnormal pulse, no user response).
 Once triggered, the system automatically:
o Sent SMS alerts to predefined emergency contacts.
o Initiated voice call alerts, delivering a preconfigured message.
o Generated in-app notifications with emergency details.
o Shared the live location (GPS coordinates) with caregivers or
emergency services.
 IoT integration further enhanced safety by automatically activating
appliances:
o Lights were turned ON to improve visibility during nighttime
emergencies.
o Smart door locks were triggered to unlock automatically,
allowing faster access for caregivers or paramedics.
 Results indicated that the average alert delivery time was between 2–5
seconds, ensuring rapid response.

Medication Reminder

 The medication module enabled users to schedule medications with

reminders at specific times.
 Push notifications were successfully delivered at scheduled times with
options to mark doses as taken or missed.
  History tracking allowed users and caregivers to review adherence
patterns.
 During testing, participants found this feature highly beneficial for
routine management of chronic diseases such as diabetes and
hypertension.
 Elderly participants appreciated the voice-based reminders, as these
were more accessible compared to text notifications.

AI Chatbot
 The AI chatbot was designed to handle basic health-related queries
and provide instant responses without needing external consultation.
 It supported voice commands and text-based interaction, ensuring
accessibility for both elderly and younger users.
 Example queries included:
o “What should I do if I feel dizzy?”
o “How often should I check my blood pressure?”
 The chatbot also guided users in app navigation, such as setting
reminders or starting health monitoring.
 During usability testing, users rated the chatbot as helpful and easy
to interact with, especially when seeking quick clarifications.

Overall System Observations

 The app maintained smooth performance with minimal lag even
when running in the background.
  Battery consumption was measured during continuous monitoring and
was within acceptable levels for a health application.
 Offline resilience: When internet connectivity was lost, the app
continued monitoring vitals locally and queued alerts for delivery
once the connection was restored.
 User feedback indicated that the interface was simple, intuitive, and
senior-friendly

4.2 TABLES, CHARTS, OR GRAPHS FOR CLARITY

Heart Rate Monitoring Results

Summary of Vitals:
 Maximum BPM: 96
 Minimum BPM: 90
  Average BPM: 93

Graph Output:
 The system plotted BPM on a line chart with time on the X-axis and
BPM on the Y-axis.
 Results showed stable vitals within the expected range.

Emergency Alert Delivery Performance

Observation:
 All alerts were delivered within 5 seconds, which is considered
acceptable for emergency scenarios.

4.3 ANALYSIS OF FINDINGS

Heart Rate Monitoring
 The system consistently measured BPM within a ±3 BPM margin of
error compared to standard medical devices such as pulse oximeters.
  Readings remained stable under normal lighting conditions, but
accuracy was slightly affected in low-light environments or when the
finger was not placed properly.
 Real-time trend visualization through graphs allowed users to monitor
short-term fluctuations and track overall heart performance.
 The local data storage ensured that sensitive health information was
not sent to third-party servers, addressing data privacy concerns.
 Findings suggest that smartphone-based PPG can serve as a cost-
effective, accessible alternative for preliminary health monitoring,
though it may not fully replace medical-grade devices.

Fall Detection
 The CNN-based model demonstrated high sensitivity and specificity
when distinguishing between normal activities (walking, sitting, lying
down) and falls.
 The model achieved strong accuracy during controlled tests, but false
positives were triggered when the phone was dropped or placed
abruptly on a surface.
 This indicates the need for further refinement, such as incorporating
multi-sensor fusion (accelerometer + gyroscope + contextual data) to
reduce errors.
 Despite minor limitations, the system proved to be a reliable tool for
fall detection, especially for elderly individuals living alone.
  Findings highlight that the model is suitable for real-world
deployment, provided continuous improvement and retraining with
larger datasets.

Emergency Alert System

 Alerts were successfully sent within 2–5 seconds, which is considered
acceptable for emergency use cases.
 Multi-channel communication (SMS, voice call, app notification)
provided redundancy, ensuring caregivers were informed even if one
channel failed.
 The integration of IoT devices such as lights and smart locks
improved environmental readiness, reducing barriers for emergency
responders.
 Live location sharing enhanced caregiver trust and response
efficiency.
 Testing confirmed that the system can operate even in low internet
connectivity areas by queuing alerts for later transmission.

Medication Reminders
 The medication reminder feature improved treatment adherence, as
users received timely push notifications.
 Elderly participants found voice reminders particularly effective
compared to text-only reminders.
 The ability to mark medications as “taken” or “missed” helped track
adherence history, enabling caregivers to identify patterns.
  This feature has potential to reduce missed dosages, which are
common in elderly patients with chronic conditions.
 Findings demonstrate that integrating simple digital reminders into a
healthcare app can significantly support preventive care and chronic
disease management.

User Feedback
 Elderly participants rated the app easy to use, praising the large
buttons, simple design, and voice interaction.
 Users reported that the “I am Okay” button was a useful safeguard
against false alerts.
 Caregivers expressed strong appreciation for the instant notifications
and live GPS tracking, which provided peace of mind.
 However, some participants noted that continuous camera use for
PPG slightly drained battery life, indicating a need for optimization in
future versions.
 Overall, the system was rated as practical, reliable, and accessible,
particularly for elderly individuals and caregivers seeking continuous
health monitoring.
 5.DISCUSSION

5.1 INTERPRETATION OF THE FINDINGS

The evaluation of Smart Health Guard confirmed that integrating AI,
IoT, and mobile health technologies into a single system can deliver
effective results in real-world scenarios.
 Heart Rate Monitoring
o The PPG-based system consistently delivered accurate BPM
readings within a small error margin.
o Graphical visualization enabled users to track long-term health
patterns, not just momentary values.
o This feature can be useful for detecting abnormal spikes or
sudden drops that might otherwise go unnoticed.
o The logging system supports the possibility of trend analysis
for future predictive healthcare models.
 Fall Detection
o The CNN model correctly identified falls in nearly all test cases.
o False positives, though rare, highlighted that the system may
mistake sudden, non-human movements for falls.
o Importantly, no critical falls went undetected, proving the
sensitivity of the model is high.
o The balance between sensitivity and specificity is crucial —
prioritizing detection over occasional false alarms ensures
safety-first performance.
 Emergency Alert System
 o Alerts were sent through multiple channels, guaranteeing
delivery.
o The inclusion of GPS tracking and IoT activation provided
caregivers with both location accuracy and environmental
support.
o Tests showed that response time was under 5 seconds, making
the system viable for real-world medical emergencies.
 Medication Reminders
o The module not only improved adherence but also helped users
develop healthy routines.
o Elderly participants highlighted that audible reminders were
particularly useful for those with poor eyesight.
o Caregivers were able to confirm adherence through the reminder
logs, ensuring better accountability.
 User Feedback
o Elderly users appreciated the minimal learning curve and
intuitive design.
o The “I am Okay” button was effective in avoiding unnecessary
panic from false alerts.
o Caregivers felt more secure knowing that they would be
notified instantly in emergencies, even if they were far away.

5.2 COMPARISON WITH PREVIOUS RESEARCH

The outcomes of this study align with existing research but also extend the
state-of-the-art in meaningful ways:
 Fall Detection
o Previous studies often required wearables (e.g., wristbands,
smartwatches). SmartHealthGuard achieved similar detection
accuracy using only built-in smartphone sensors, reducing
dependency on costly devices.
o This makes the system more scalable in low-income or rural
settings.
 Heart Rate Monitoring
o Earlier smartphone PPG studies validated feasibility but did not
integrate features like trend visualization, historical logs, and
anomaly alerts. SmartHealthGuard addresses these gaps.
 mHealth Applications
o Most existing apps are feature-specific — either medication
reminders, vitals tracking, or chatbots. SmartHealthGuard’s
integrated ecosystem reduces the need for multiple
applications.
 IoT-enabled Care
o Prior research emphasized IoT’s potential but often required
complex home automation setups. This project shows that
even basic IoT devices (lights, doors) can significantly improve
emergency accessibility.
 Elder-Friendly Design
o While many health apps are designed for younger, tech-savvy
users, SmartHealthGuard demonstrates that simplified HCI
 (large icons, voice commands, intuitive navigation) significantly
improves adoption by older users.

5.3 IMPLICATIONS OF THE STUDY

1. Practical Implications
o Provides real-time intervention that can prevent serious health
complications or deaths.
o Offers cost-effective healthcare access in regions with limited
medical infrastructure.
o Reduces caregiver stress by offering continuous, remote
monitoring.
o Supports aging-in-place, allowing elderly individuals to live
independently while staying safe.
o Encourages preventive healthcare, potentially lowering
hospital admissions.
2. Theoretical Implications
o Confirms that AI-driven fall detection can be deployed on
mobile devices without external wearables.
o Strengthens the case for mHealth as a disruptive healthcare
model, blending AI, IoT, and HCI.
o Demonstrates that multi-functionality within one platform
improves adoption compared to single-feature apps.
o Validates HCI principles in healthcare design, especially the
role of voice-enabled interaction in accessibility.
5.4 LIMITATIONS OF THE RESEARCH

Although the Smart Health Guard application achieved promising

results, a few limitations were observed during testing:
1. False Positives in Fall Detection – The CNN model occasionally
misclassified sudden non-human movements (such as phone drops) as
falls, indicating the need for further refinement.
2. Local Data Storage Only – Health data was stored locally for
privacy reasons, but this limited remote monitoring and multi-device
synchronization.
 6. CONCLUSION

6.1 SUMMARY OF KEY FINDINGS

This project developed and evaluated SmartHealthGuard, an AI-powered,
voice-activated healthcare application designed for real-time health
monitoring and emergency response. The findings demonstrate that
integrating AI, IoT, and mHealth technologies in a single system can
significantly improve healthcare accessibility and responsiveness,
particularly for elderly individuals and patients with chronic conditions.
Key outcomes include:
 Heart Rate Monitoring: The smartphone-based PPG system
accurately measured heart rate within a ±3 BPM margin of error,
validated against medical-grade devices. Real-time charts and history
logs enhanced usability.
 Fall Detection: The CNN-based algorithm successfully identified
falls in test cases, ensuring high sensitivity. False positives occurred
during non-human movements (e.g., dropping the phone), but no
actual falls went undetected.
 Emergency Alert System: Alerts reached caregivers within 2–5
seconds via SMS, calls, and notifications. IoT integration (lights,
doors, alarms) improved emergency readiness.
 Medication Reminders: Timely reminders improved adherence to
treatment schedules. Voice notifications were highly effective for
elderly users.
  User Feedback: Participants rated the app as easy to use, accessible,
and reliable, especially due to its voice interaction and simplified
interface. Caregivers valued instant alerts and live GPS tracking.
These results confirm that SmartHealthGuard is a cost-effective, practical,
and user-friendly solution for personal healthcare management.

6.2 RECOMMENDATIONS FOR FUTURE RESEARCH

To strengthen the system and broaden its applicability, the following
directions are recommended:
1. Enhanced Fall Detection – Incorporating larger datasets and multi-
sensor fusion (accelerometer, gyroscope, contextual data) to
minimize false positives.
2. Expanded Vitals Monitoring – Adding functionalities for blood
pressure, oxygen saturation (SpO₂), and ECG monitoring using
smartphone-compatible sensors.
3. Cloud Integration – Implementing secure, encrypted cloud storage
to enable long-term record-keeping and remote access by healthcare
professionals.
4. Predictive Analytics – Leveraging AI to detect patterns in health
data and issue early warnings for potential emergencies.
5. Large-Scale Clinical Trials – Conducting studies with diverse
populations to validate accuracy and usability in real-world
healthcare environments.
6. Offline Functionality – Improving resilience for low-connectivity
areas, ensuring uninterrupted safety features.
6.3 PRACTICAL IMPLICATIONS OF THE RESULTS
The project carries significant real-world applications:
 For Elderly Users: Offers independence and confidence by ensuring
help is always available during emergencies.
 For Caregivers: Provides peace of mind through instant alerts, live
location sharing, and adherence tracking.
 For Healthcare Systems: Encourages preventive care and reduces
avoidable hospital admissions, easing the burden on healthcare
infrastructure.
 For Technology Adoption: Demonstrates that smartphones alone,
without expensive wearables, can be used for AI-powered health
monitoring.
 For Accessibility: Validates the importance of voice-based
interaction and elder-friendly design in healthcare apps, ensuring
wider adoption.
REFERENCES
AI-Based Healthcare Monitoring System A study on how artificial
intelligence can be used for continuous health monitoring and anomaly
detection in patients. (IEEE Xplore, 2021)

Wearable IoT for Healthcare Monitoring Research on the integration of

IoT-enabled devices for monitoring vital signs like blood pressure, heart
rate, and blood sugar levels in real-time. (Springer, 2020)

Artificial Intelligence in Fall Detection for Elderly Research on AI-

based fall detection systems that help elderly individuals by automatically
alerting caregivers and emergency services. (Elsevier, 2022)

Smartphone-Based Photoplethysmography (PPG) for Heart Rate

MonitoringA study demonstrating the use of smartphone cameras and
flashlight for accurate, low-cost heart rate monitoring using
photoplethysmography. (Nature Scientific Reports, 2019)

Voice-Activated Health Assistants Research on integrating natural

language processing and AI chatbots for elder-friendly healthcare support
and real-time query resolution. (ACM Digital Library, 2021)
 CODING

1 FALL DETECTION (CNN MODEL WITH SMARTPHONE

SENSORS)
import tensorflow as tf
import numpy as np

# Load trained TensorFlow Lite model

interpreter = tf.lite.Interpreter(model_path="fall_detection.tflite")
interpreter.allocate_tensors()

# Define input/output details

input_index = interpreter.get_input_details()[0]['index']
output_index = interpreter.get_output_details()[0]['index']

def detect_fall(sensor_values):
# Preprocess accelerometer + gyroscope data
input_data = np.array(sensor_values, dtype=np.float32).reshape(1, -1)

# Run model inference

interpreter.set_tensor(input_index, input_data)
interpreter.invoke()

prediction = interpreter.get_tensor(output_index)
 if prediction[0][0] > 0.8: # Fall threshold
print("Fall Detected! Triggering Alert...")
trigger_emergency_alert()
else:
print("No Fall Detected")

def trigger_emergency_alert():
# Placeholder for SMS/voice call API
print("Emergency Alert Sent to Caregivers")

2 HEART RATE MONITORING (PPG USING CAMERA +

FLASHLIGHT)
import 'package:camera/camera.dart';
import 'dart:math';

CameraController? controller;
List<int> intensityValues = [];

Future<void> startHeartRateMeasurement() async {

controller = CameraController(
cameras.first, ResolutionPreset.medium,
enableAudio: false);

await controller!.initialize();
controller!.startImageStream((CameraImage image) {
 // Extract red channel intensity
int avgRed = image.planes[0].bytes.reduce((a, b) => a + b) ~/
image.planes[0].bytes.length;
intensityValues.add(avgRed);

if (intensityValues.length > 100) {

calculateBPM(intensityValues);
intensityValues.clear();
}
});
}

void calculateBPM(List<int> values) {

// Apply signal processing to estimate BPM
double frequency = fft(values); // Fast Fourier Transform
int bpm = (frequency * 60).round();
print("Heart Rate: $bpm BPM");
}
 SUPPLEMENTARY MATERIALS

1. QUESTIONNAIRE (USER FEEDBACK TESTING)

Section A – User Profile
1. Age: ___
2. Gender: ___
3. Do you have chronic conditions? (e.g., diabetes, hypertension, heart
disease)
Section B – Usability
1. Was the app easy to install and set up? (Yes/No)
2. Was the interface clear and readable? (1–5 scale)
3. Did you find voice interaction useful? (1–5 scale)
Section C – Feature Effectiveness
1. How useful was the fall detection feature? (1–5 scale)
2. Were medication reminders delivered on time? (Yes/No)
3. Did emergency alerts reach caregivers quickly? (Yes/No)
4. How reliable was the heart rate monitoring? (1–5 scale)
Section D – Overall Feedback
1. Would you recommend Smart Health Guard to others? (Yes/No)
2. Suggestions for improvement: _______________________