Emotion recognition in blurred images with local features and machine learning

Kristensen, Rasmus Lyngby

4. term

Vision, Graphics and Interactive Systems, Master

2014

2014-06-03

130 pages

Facial Expression Recognition (FER) has important extensions to the development of the next generation of Human Machine Interfaces (HMI). In this project, it was proposed to solve the problem of FER in blurred images by using the blur invariant local feature descriptor Local Frequency Descriptor (LFD). Further, it was proposed to use Spectral Clustering (SC) to reduce the dimensionality of local features for FER. The recognition accuracy of LFD was com- pared against that of the blur invari- ant local feature descriptor Local Phase Quantization (LPQ) and the popular local feature descriptor Local Binary Patterns. The recognition accuracy of SC was com- pared against that of Principal Component Analysis (PCA). All recognition accura- cies was measured using Support Vector Machines (SVM). It was found, that LPQ seems to actually provide better recognition accuracy than LFD. Further, it was found that SC in general provides similar or slightly better results than PCA. However, the SC method is more computational expensive.

Blur ; Computer Vision ; Machine Learning ; Local Features ; Local Binary Patterns ; Local Directional Pattern variance ; Local Phase Quantization ; Local Frequency Descriptor ; Spectral Clustering ; Principal Component Analysis ; Segmentation ; Face ; Facial Expression ; Emotion ; Recognition ; Robustness ; Blur robustness

Download
View record in AAU Student Projects