Skip to content
All work

Machine Learning · 2017

Stress Analyser

Facial recognition-based system to detect and classify stress levels.

Illustrative data sketch, not a project screenshot

Overview

Stress Analyser estimates how stressed someone is from a photo of their face. It's written in Python: Haar Cascade classifiers find the face, the program extracts features linked to stress from the expression, Principal Component Analysis (PCA) reduces them, and a Support Vector Machine (SVM) classifies the result. It detected stress with over 88% accuracy, using nothing but the image, so no sensors touch the person.

What I did

  1. 01Detected faces with Haar Cascade classifiers
  2. 02Extracted stress-related features from facial expressions
  3. 03Used PCA to reduce the number of features
  4. 04Classified stress levels with an SVM
  5. 05Reached over 88% accuracy
Next projectDiverse Supplier Chatbot