https://github.com/mnxtr/cloudbasedbiometric-sensors
Research on AI based ekg interpretation of myocardial infarction using multiple neural networks.
https://github.com/mnxtr/cloudbasedbiometric-sensors
ad8232 ad8232ecg arduino arduino-sketch cpp ecg-classification ecg-qrs-detection iot ipynb
Last synced: about 2 months ago
JSON representation
Research on AI based ekg interpretation of myocardial infarction using multiple neural networks.
- Host: GitHub
- URL: https://github.com/mnxtr/cloudbasedbiometric-sensors
- Owner: mnxtr
- License: cc0-1.0
- Created: 2023-03-10T15:47:16.000Z (about 3 years ago)
- Default Branch: main
- Last Pushed: 2025-11-05T17:31:59.000Z (4 months ago)
- Last Synced: 2025-11-05T19:37:27.936Z (4 months ago)
- Topics: ad8232, ad8232ecg, arduino, arduino-sketch, cpp, ecg-classification, ecg-qrs-detection, iot, ipynb
- Language: Jupyter Notebook
- Homepage:
- Size: 38.1 KB
- Stars: 1
- Watchers: 0
- Forks: 1
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Security: SECURITY.md
Awesome Lists containing this project
README
# CSE498R Directed Research | Cloud-based Biometric Sensor
[Tasks](https://www.notion.so/ab800a73cf014937b95d28b0d7752da7)
[Task list](https://www.notion.so/ab800a73cf014937b95d28b0d7752da7)
To verify that the heart rate monitor is working as expected, [open the serial monitor](https://learn.sparkfun.com/tutorials/terminal-basics/arduino-serial-monitor-windows-mac-linux)
at **9600 baud**
## interfacing with arduino and AD8232
### **/src/sketch.ino**
```cpp
#include
#include
// Define the AD8232 module
AD8232 heartRate;
void setup()
{
Serial.begin(9600); // Initialize serial communication
heartRate.begin(); // Initialize the AD8232 module
}
void loop()
{
float heart_rate = heartRate.getHeartRate(); // Read the heart rate value
Serial.println("Heart rate: " + String(heart_rate) + " bpm"); // Print the heart rate value
delay(1000); // Wait for 1 second
}
```
. You should see values printed on the screen. Below is an example output with the sensors connected to the forearms and right leg. Your serial work should spike between +300/-200 around the center value of about ~500.
## List of hardware required
- Arduino Rev3 atmega328 + Wi-Fi → as MCU
- [AD8232 Heart Sensing Monitor](https://www.analog.com/en/products/ad8232.html)
- Breadboard
- Breadboard connector
- Electrode Sensor Surface pad → Disposable
- Pulse Sensor Module.
- `EEG Muscle Sensor.`
---
## **Software**
- Arduino IDE
- [Arduino Cloud - native](https://cloud.arduino.cc/home/?get-started=true)
- Fritzing
- [Excali Draw](https://excalidraw.com/)
---
> Serial Plotter
>
Initial Prototype
First Test Case Requires Finetuning
## Arduino Rev3 atmega328 + AD8232 Connection.
Wiring
-
- Cable Color Signal **Black** RA (Right Arm ) | **Blue** LA (Left Arm) | **Red** RL (Right Leg)
Enhancement+
Arduino layout with 7 Segment Led
Research design and problem addressing
-----------------
- A research problem is a specific issue, contradiction, or gap in existing knowledge that needs to be addressed.
- Identifying a research problem involves recognizing something problematic that requires attention, explanation, or a solution.
- It is essential to have a well-defined research problem to guide the research process and to contribute to the field of study.
- The process of defining a research problem includes:
- Identifying a broad topic of interest.
- Conducting a literature review to understand the current state of knowledge.
- Narrowing down to a specific issue that is under-explored or controversial.
- Formulating the problem into a clear, concise statement or question.
- Types of research problems include:
- Descriptive: Documenting certain phenomena or situations.
- Exploratory: Investigating a topic to generate new insights or hypotheses.
- Explanatory: Understanding the causes or effects of certain phenomena.
- Predictive: Anticipating future occurrences based on current trends or data.
- Evaluative: Assessing the effectiveness of interventions, programs, or policies.
- Defining a research problem is crucial for:
- Setting the direction and scope of the study.
- Ensuring the research is focused and manageable.
- Avoiding duplication of existing knowledge.
- Making a meaningful contribution to the academic field or practical application.
- For more detailed guidance on defining a research problem, resources such as Scribbr's articles can provide step-by-step instructions and examples.
## Scopes
---
## BPNN Model Integration
The repository now includes comprehensive mathematical formulations for integrating a Backpropagation Neural Network (BPNN) with the EKG system.
### Mathematical Formulations Documentation
See [BPNN_MATHEMATICAL_FORMULATIONS.md](BPNN_MATHEMATICAL_FORMULATIONS.md) for complete mathematical documentation including:
#### Core BPNN Mathematics
- **Forward Propagation**: $z^{[l]} = W^{[l]} a^{[l-1]} + b^{[l]}$
- **Activation Functions**: Sigmoid, Tanh, ReLU, Softmax (with derivatives)
- **Loss Functions**: Binary/Categorical Cross-Entropy, MSE
- **Backpropagation**: Complete gradient computation formulas
- **Parameter Updates**: Gradient Descent, Momentum, Adam optimizer
#### Signal Processing
- **Normalization**: Min-Max and Z-Score formulas
- **Filtering**: Moving Average and Butterworth Bandpass filters
- **Feature Extraction**: HRV metrics (SDNN, RMSSD, pNN50)
#### EKG Synthesis Model
Mathematical model for generating synthetic EKG signals:
$$\text{ECG}(t) = \sum_{i \in \{P, Q, R, S, T\}} A_i \exp\left(-\frac{(t - t_i)^2}{2\sigma_i^2}\right)$$
### Jupyter Notebooks
#### 1. Data Loading and Preprocessing (`dataloading.ipynb`)
- Signal normalization implementations
- EKG filtering techniques
- HRV feature extraction
- Window-based segmentation for neural network input
- Complete preprocessing pipeline with visualizations
#### 2. BPNN Implementation (`randomdatasetgenerator.ipynb`)
- Complete BPNN implementation from scratch
- Synthetic EKG dataset generation (normal and abnormal patterns)
- Training and evaluation pipeline
- Multi-class classification (Normal, Tachycardia, Bradycardia, Arrhythmia)
- Performance visualization and analysis
### Quick Start
To use the BPNN model:
```python
# Install required dependencies
pip install numpy pandas matplotlib scipy scikit-learn
# Run the validation script
python3 validate_formulations.py
# Open Jupyter notebooks
jupyter notebook dataloading.ipynb
jupyter notebook randomdatasetgenerator.ipynb
```
### Model Architecture
The implemented BPNN uses the following architecture:
- **Input Layer**: 250 neurons (EKG window samples)
- **Hidden Layer 1**: 128 neurons (ReLU activation)
- **Hidden Layer 2**: 64 neurons (ReLU activation)
- **Hidden Layer 3**: 32 neurons (ReLU activation)
- **Output Layer**: 4 neurons (Softmax activation for 4 classes)
**Total Parameters**: ~54,000 trainable parameters
### Classification Classes
The model classifies EKG signals into four categories:
1. **Normal**: Regular heart rhythm (60-100 bpm)
2. **Tachycardia**: Fast heart rate (>100 bpm)
3. **Bradycardia**: Slow heart rate (<60 bpm)
4. **Arrhythmia**: Irregular heart rhythm
---