Smart Riding Helmet

Machine Learning module for the AI-Based Smart Riding Helmet system. This module analyzes rider behavior using alcohol detection and drowsiness monitoring data to classify rider safety risk levels.

---

Overview

The Smart Riding Helmet uses sensor data collected from the MQ3 Alcohol Sensor and Eye Blink Sensor to evaluate rider safety. A Support Vector Machine (SVM) model is trained to classify riders into different risk categories based on their behavior and sensor readings.

The trained model can be integrated with the ESP32-based monitoring system and dashboard to provide intelligent safety recommendations and alerts.

---

Objective

The objective of this module is to:

• Detect unsafe riding conditions.
• Analyze alcohol consumption levels.
• Monitor drowsiness indicators.
• Classify rider risk levels.
• Support real-time rider safety monitoring.

---

Dataset Features

The dataset contains rider behavior and sensor readings collected during testing.

Feature	Description
alcohol_level	MQ3 sensor alcohol reading
blink_rate	Number of eye blinks per minute
eye_closure_duration	Eye closure duration in milliseconds
speed	Vehicle speed during monitoring
risk_status	Target label

---

Risk Classes

Class	Description
Safe	Normal riding condition
Caution	Moderate risk detected
Danger	High-risk riding condition

---

Feature Engineering

The preprocessing pipeline performs:

• Duplicate removal
• Missing value handling
• Drowsiness score calculation
• Alcohol risk categorization
• Speed risk categorization
• Combined risk score generation
• Target label encoding

Generated Features:

• drowsiness_score
• alcohol_risk
• speed_risk
• combined_risk
• risk_status_encoded

---

Machine Learning Model

Support Vector Machine (SVM)

Support Vector Machine was selected because it performs well on structured sensor data and provides reliable classification for rider safety prediction.

Configuration

SVC(
    kernel='rbf',
    C=1.0,
    gamma='scale',
    probability=True,
    random_state=42
)

---

Project Structure

model_training/
│
├── rider_dataset.csv
├── processed_dataset.csv
├── feature_engineering.py
├── svm_training.ipynb
├── train_svm.py
├── svm_model.pkl
├── scaler.pkl
├── confusion_matrix.png
├── accuracy_curve.png
├── requirements.txt
└── README.md

---

Technologies Used

• Python
• Pandas
• NumPy
• Scikit-Learn
• Support Vector Machine (SVM)
• Matplotlib
• Joblib
• Jupyter Notebook

---

Installation

Install required packages:

pip install -r requirements.txt

---

Training the Model

Step 1: Feature Engineering

python feature_engineering.py

Output:

processed_dataset.csv

Step 2: Train SVM Model

python train_svm.py

Outputs:

svm_model.pkl
scaler.pkl
confusion_matrix.png
accuracy_curve.png

---

Model Evaluation

The model is evaluated using:

• Accuracy
• Precision
• Recall
• F1 Score
• Confusion Matrix

Generated visualizations:

• confusion_matrix.png
• accuracy_curve.png

---

Sample Prediction Workflow

MQ3 Alcohol Sensor
        +
Eye Blink Sensor
        +
Vehicle Speed
        ↓
Feature Engineering
        ↓
SVM Classifier
        ↓
Risk Prediction
        ↓
Safe / Caution / Danger

---

Integration with Smart Riding Helmet

The trained SVM model can be integrated with:

• ESP32 Controller
• Smart Riding Helmet Dashboard
• Rider Alert System
• GPS Tracking Module

to provide intelligent rider safety monitoring and decision support.

---

Future Improvements

• Real-time model inference
• Larger rider dataset
• GPS-aware risk prediction
• Deep learning-based drowsiness detection
• Mobile application integration
• Cloud-based analytics dashboard

---

License

This project is developed for educational, research, and portfolio purposes.

---

Author

Janani V

Machine Learning | IoT | Embedded Systems

GitHub: https://github.com/Janviswa

LinkedIn: https://www.linkedin.com/in/jananiv05/