Facial Expression Analysis Based on LBPs Feature

This project was completed as part of my Master’s degree in Computer Science. In 2009, my advisor was visiting a research group at UCSD, where they were exploring the use of local binary patterns in both 2D and 3D sequences as features for facial expression classification, including micro-expressions and their combinations. I was able to align my research with that group and contribute my insights to the project. Additionally, I discovered a simple method to reduce data dimensions without compromising accuracy.

I worked on this project independently, coding everything from scratch, including the feature extraction code and linear SVM training code. I used C++ and C# to integrate all components, putting my knowledge of software engineering into practice. The entire system included core algorithms coded in C++ and module interfaces designed and implemented using C#. These two major parts were connected using C++.net. The GUI of the final software could automatically adjust based on the types of models loaded.

Here is a test video recorded in the early stages of the project: