Face Expression Recognition (FER) has immense popularity in research community due to its widespread applicability in different applications. However, the currently available expression descriptors are suffering from several problems like noises, distortions and weak edges. To sort out these issues, this paper proposes an Edge Aware Directional Coding (EADC) that efficiently encodes the prominent expression related features. Initially, EADC finds out the gradients based on Kirsch Compass Mask in eight directions. Then EADC encode each pixel in facial image based on the Gradient Support of its neighbor pixels. Directional Edge Weight (DEW) is measured as support based on the gradient magnitude and gradient direction of edges passing through the neighbor pixels. EADC encode all the texture primitives such as corners and edges such that the recognition system can recognize expressions accurately independent of subjects. For experimental validation, JAFFE and CK+ datasets are used and the performance is measured through recognition accuracy. The obtained results show that the EADC outperforms all the earlier state-of-the-art methods.