An Efficient Hybrid Technique of Feature Extraction for Facial Expression Recognition

A single technique can extract only some of feature of facial expression image, but hybrid methods of different techniques have all combined advantage for feature extraction. The author has proposed a hybrid feature extraction technique based on the Discrete Cosine Transform, Wavelet transform, Gabor filter and Gaussian distribution. In the proposed scheme DCT and Discrete wavelet transform, Gabor filter and Gaussian distribution based feature extraction technique is applied separately and optimum features from each method is merged in a feature vector. Algorithms are implemented in Matlab and JAFEE dataset are used for experiment with ratio 70/30 of training/testing with adaboost classifiers for seven different facial expressions. The classification result of the proposed work is compared with DCT based feature extraction technique, Gabor Filter based feature technique, DWT based feature extraction technique and Gaussian distribution based technique.