Automatic recognition and interpretation of human emotions are becoming an integral part of intelligent products and services, which can lead to a breakthrough in domains such as healthcare, marketing, security, education and environment. This leads us towards Facial Expression Recognition (FER) systems, whose main objective is to detect an expressed emotion and recognize the same. The proposed work introduces an FER system, which models the relationship between human facial expression and corresponding induced emotion for static images by extracting shape and appearance features. Automatic interpretation of facial expressions from static images is very challenging, as information available in a static image is less when compared with image sequences. Proposed method incites efficient extraction of shape and appearance features from still images for varying facial expressions using Histogram of Oriented Gradient (HOG) features, and Support Vector Machine (SVM) is employed for classification. The proposed work is implemented on Cohn-kanade data set for six basic expressions (happy, sad, surprise, anger, fear and disgust). Results show that identification of human expressions from static images has a better detection rate when shape and appearance features are used rather than texture or geometric features. Use of HOG for feature extraction makes the FER system robust, and real time implementation is made much easier.