Live AI Cardiac Monitoring · GNIOT CSE-IoT · 2025-26

Predict Cardiac Events

Before They Happen

Heartlyf fuses TinyML edge intelligence, ESP32 wearable hardware and real-time Firebase cloud to continuously monitor ECG, detect arrhythmias and dispatch GPS-enabled emergency alerts in under one second.

▶ Start Monitoring Free Explore the Project
LIVE ECG 72 BPM Normal Sinus Rhythm Hover to inspect mV values
99.9%
Monitoring Uptime
<1s
Alert Latency
AI Detection Layers
24/7
Real-time Monitoring

Built for Real-World Medical Monitoring

8 cutting-edge features that make Heartlyf the most advanced IoT cardiac monitoring platform available.

🧠

AI Arrhythmia Detection

TinyML model detects 5 types of cardiac arrhythmias with >95% accuracy in real-time.

⚙️

Edge AI Processing

On-device inference using ESP32 with quantized TensorFlow Lite — no cloud latency.

📍

GPS Emergency Location

NEO-6M GPS sends precise coordinates with every emergency alert via Telegram.

📈

Real-time ECG Monitoring

AD8232 sensor captures 250Hz ECG signals with 12-bit ADC precision continuously.

📱

Web App Integration

Monitor vitals from any device via responsive PWA with full support.

📋

Doctor Reports

Auto-generated PDF medical reports with ECG strips, HRV analysis and AI notes.

🔔

Family Alerts

Instant Telegram, SMS and push notifications to family during cardiac events.

6-Layer Architecture

End-to-end pipeline from IoT sensor to emergency response.

01
📡
Sensors
AD8232 ECG
MAX30105 SpO2
NEO-6M GPS
02
Signal Processing
Bandpass filter
Noise removal
250Hz sampling
03
🧠
TinyML Model
TF Lite on ESP32
<100ms inference
5-class classifier
04
🔲
Edge AI
ESP32 240MHz
Local prediction
WiFi sync
05
☁️
Cloud Analytics
Firebase RT DB
Deep Learning
Dashboard sync
06
🚨
Emergency Response
Telegram + SMS
GPS location
Auto call
📊 Real-time Data Flow Pipeline
📉
ECG Capture
AD8232 · 250Hz
🔊
Filter & Clean
0.5–40Hz BPF
📐
Feature Extract
PQRST · HRV
🤖
AI Classify
TF Lite · <100ms
📤
Cloud Sync
Firebase RTDB
📱
Dashboard
Real-time UI

🔧 Software Stack

🧠

TensorFlow Lite

Quantized INT8 model for ESP32 edge inference

🔥

Firebase Realtime DB

Sub-second cloud sync for ECG vitals and alerts

⚛️

React + Chart.js

Live ECG visualization and patient dashboards

📡

Telegram Bot API

Instant emergency notifications with GPS coordinates

🐍

Python / Scikit-learn

MIT-BIH training pipeline and model evaluation

🔌 IoT Hardware Schematic ESP32 Core
ESP32 240MHz · TinyML WiFi + BT 12-bit ADC AD8232 ECG Sensor 250Hz · 12-bit MAX30105 SpO2 + HR PPG I2C Interface NEO-6M GPS Module UART · ±3m SD Card Local Storage SPI · 32GB Firebase RTDB Cloud WiFi 802.11 LiPo 2000mAh 3.7V Battery USB-C Charge

🤖 AI Classification Accuracy MIT-BIH Arrhythmia Database · 48 recordings

99.8%
Normal Sinus
Baseline rhythm
98.5%
Atrial Fibrillation
Irregular rhythm
97.2%
Ventricular Tachy
Rapid VT
96.8%
Premature Beat
PVC / PAC
95.4%
Bundle Branch Block
Conduction delay

Physical Hardware Stack

🔲

ESP32 Microcontroller

Main controller with dual-core 240MHz CPU, WiFi & Bluetooth. Runs TinyML edge AI model locally.

📉

AD8232 ECG Sensor

Dedicated cardiac monitor IC with low-power op-amp. Captures clean ECG waveforms at 250Hz.

❤️

MAX30105 SpO2 Sensor

Optical heart-rate and SpO2 sensor using PPG technology. Accuracy ±2% SpO2.

🔋

LiPo Battery

2000mAh rechargeable LiPo with USB-C charging. Up to 48h continuous monitoring.

5-Stage AI Pipeline

01
📡

Signal Acquisition

AD8232 ECG + MAX30105 SpO2 capture physiological signals at 250Hz sampling rate

02

Signal Preprocessing

Bandpass filter (0.5–40Hz), baseline wander removal, noise reduction on ESP32

03
🧠

AI Prediction

TensorFlow Lite model classifies ECG rhythm in <100ms on ESP32 edge device

04
☁️

Cloud Analytics

Data synced to Firebase Realtime DB. Deep learning model on cloud runs secondary analysis

05
🚨

Emergency Response

Telegram + SMS alerts with GPS coordinates sent instantly to emergency contacts

MIT-BIH Training Data

Cardiac analysis engine trained using the MIT-BIH Arrhythmia Database — gold-standard ECG dataset provided by PhysioNet.

Dataset Details

SourcePhysioNet
DatasetMIT-BIH Arrhythmia Database
PublicationMoody & Mark, IEEE EMB, 2001
Sampling Rate360 Hz
Records48 annotated ECG recordings

Key Features

  • 48 half-hour ECG recordings from 47 subjects
  • 360 Hz sampling with 11-bit resolution
  • Expert-annotated arrhythmia beat classifications
  • Clinical-grade ECG signals for deep learning
Moody GB, Mark RG. "The impact of the MIT-BIH Arrhythmia Database." IEEE Engineering in Medicine and Biology Magazine, 2001.
View on PhysioNet ↗

Built by Innovators

AI-Powered Cardiac Monitoring System — Senior Capstone Project, GNIOT 2025-26

A

Anand Asati

Frontend Developer
B.Tech IoT, Final Year
D

Devant Pal

Cloud, Database & AI/ML
B.Tech IoT, Final Year
G

Gajendra Kumar

Backend Developer
B.Tech IoT, Final Year
M

Md Maaz Yazdani

IoT & Hardware
B.Tech IoT, Final Year
Project
AI-Powered IoT-Based Autonomous Cardiac Arrest Prediction System
Supervisor
Mr. Neelash Shell
Department
Computer Science & Engineering – IoT
Academic Year
2025–2026