Recognition of Face Expression using Color Space

Face expression recognition can be stated as „identifying the expression of an individual from images of the face‟. Most of the existing systems of facial expression recognition focus on gray scale image features. This paper describes the novel approaches for effectively recognizing the facial expressions. In facial expression recognition (FER) framework, initially the face region of the image is detected using Ada boost learning algorithm. The RGB color image is transformed to another color space (YCbCr) to further improve face recognition performance. From this, the features are extracted by using Log-Gabor filter which results in a large amount of feature arrays. Furthermore, to improve the classification accuracy, the most desirable features are selected by using mutual information quotient (MIQ) technique. At last the multiclass linear discriminant analysis (LDA) classifier is used to classify these features. The experimental results evaluate the performance of using this framework under varying illumination and in low resolution images. Keywords— Log-Gabor filter, linear discriminant analysis, mutual information quotient, color space, facial expression recognition.