The "Facial Expressions Recognition Using the Emotiv EPOC Headset" project developed for the "Sensor Signal Processing" course within the EMECS Masters
- Date: March 2018
- Purpose: The purpose of this project is to develop a program which recognizes in real-time the Neutral state and 5 facial expressions (left wink / blink, right wink / blink, strong blink, open mouth, full mouth) using the 14 channels data provided by the Emotiv EPOC Headset
- Programming Language: Python
- Team:
- Vitor Ribeiro Roriz
- Alexandru Cohal
- Inputs:
- The 14 channels data provided by the Emotiv EPOC Headset (monitoring the electrical activity of the muscle tissue)
- Outputs:
- The facial state (neutral, left wink / blink, right wink / blink, strong blink, open mouth, full mouth)
- Solution:
- The classical steps were followed: data acquisition, data preprocessing, feature extraction and classification
- A comparison between multiple calssifiers was performed:
- Decision Tree classifier based on thresholds
- Multi-Layer Perceptron
- Support Vector Machine
- K-Nearest Neighbours
- For more information about the solution, implementation, results, conclusions and improvements see this document
- Results:
- Real-Time classification of the previously specified facial expression was sucessful with a period of 0.07 seconds
- The best classifier was K-Nearest Neighbours (K = 11)
- For more information about the solution, implementation, results, conclusions and improvements see this document