Facial Gesture Classifier for Children with Cerebral Palsy
Undergraduate Thesis - PRISM Lab, Holland Bloorview Kids Rehabilitation Hospital
For my undergraduate thesis, I worked with Dr. Tom Chau in PRISM Lab at Holland Bloorview Kids Rehabilitation Hospital to develop access technology for children with cerebral palsy. I used MATLAB to develop a facial gesture classifier to enable children to interact with computes and devices. The classifier uses supervised learning through a Hidden Markov Model to recognize smiling, mouth opening, neutral, and miscellaneous gestures. Using hierarchical classification, it achieved accuracy up to 98% when tested on database videos.
My thesis was awarded the Centennial Senior Project Award ($500). My first-authored paper for this research was recently submitted, and preparation for a patent for this technology is currently underway.
Contributions
- Conducted independent research for cerebral palsy rehabilitation
- Developed software pipeline and wrote integration code in MATLAB
- Designed a hierarchical classifier architecture to significantly improve accuracy
- First-authored a journal paper, recently submitted for review
Skills
- Independent research
- Programming in MATLAB
- Image processing
- Computer vision
- Graphic design
Check out the summary presentation of this research here!