Stress Detection System
Final Year Project Presentation
Ankit Thakur & Pravesh Kumar
Galgotias University, 2025
Abstract
• Stress detection using GSR and EDA signals.
• ML models: Decision Tree, SVM, etc.
• Goal: Real-time stress level prediction from
physiological data.
System Overview
• Problem: High stress levels in working
individuals.
• Goal: Early detection of stress using sensors
and ML.
• Approach: Use GSR data + ML classification.
Technologies Used
• Hardware: Arduino Uno, GSR Sensor.
• Software: Python, Arduino IDE.
• ML Algorithm: Decision Tree Classifier.
System Design
• Data Collection → Preprocessing → Model
Training.
• Prediction: Uses trained ML model on new
data.
• Architecture includes sensor, controller,
storage, model.
Feature Engineering
• Normalization, Label Encoding, PCA.
• Outlier detection and smoothing for signal
clarity.
• Feature extraction critical for ML
performance.
Algorithm Comparison
• Decision Tree: 89% Accuracy
• SVM: 91%, Random Forest: 93%
• Evaluation metrics: Precision, Recall, ROC.
Applications
• Students, Teachers, Healthcare Workers.
• Real-time stress monitoring.
• Useful for mental health management.
Ethical Considerations
• Data privacy and anonymization.
• User consent is mandatory.
• Compliance with GDPR and Indian laws.
Case Study
• Subject A: Normal stress variation.
• Subject B: High stress before deadlines.
• Visualization of stress levels over time.
Limitations
• Sensor variability and noise.
• User data diversity needed.
• Hardware wearability challenges.
Conclusion
• GSR-based detection is ~91% effective.
• Machine learning improves prediction
accuracy.
• Future: multi-sensor systems and deep
learning models.

Stress_Detection_System_Presentation.pptx