Emotion recognition from facial expressions
Recognizing facial emotions is a fundamental aspect of interpersonal communication. People with diseases like Autism, Alzheimer's disease or Parkinson's disease have impairment to understand other people's emotions. In order to help people who are unable to visualize people's fac...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2015
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/63479 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Recognizing facial emotions is a fundamental aspect of interpersonal communication. People with diseases like Autism, Alzheimer's disease or Parkinson's disease have impairment to understand other people's emotions. In order to help people who are unable to visualize people's facial emotions during their face to face communication, a need of real-time emotion recognizer is required. The objective of the project is to study some selected existing facial emotion recognition algorithms and implement the most suitable algorithm. The emotion recognition software is written in Python language with OpenCV as the main library to run image processing tools in the program. Haar-like classifier is used to detect face, mouth, and eyes region. After identification of ROI (Regions of Interest), features extraction is required for the software to identify emotions. Shi-Tomasi corner detector is used to collect distance between corners of the lip. Other than corners of the lip, teeth area is also computed in order to help identifying emotions. For the eyes region, Hough circle transform is utilized to identify large eye-opening. From all the features extracted from the image, four basic emotions can be identified by the software. They are neutral, happy, fear, and surprise. From the experiment where participant’s expression maintains the same expression under three minutes for three times run, the accuracy table can be created. The neutral has the highest rate of accuracy with 100% correctness. The happy emotion has the accuracy of 75.33%. Fear emotion has an accuracy of 74.84%. Finally, surprise emotion has second highest rate of 93.6% accuracy. |
|---|