Abstract: The success of a good facial expression recognition system depends on the facial feature descriptor. Features extracted from local region are widely used for facial expression recognition due to their simplicity but the long feature vector length produces by them makes the overall system slow for recognition. This study presents a unique local facial feature descriptor, the Local Arc Pattern (LAP) for facial expression recognition. Feature is obtained from a local 5x5 pixels region by comparing the gray color intensity values surrounding the referenced pixel to formulate two separate binary patterns for the referenced pixel. Each face is divided into equal sized blocks and histograms of LAP codes from those blocks are concatenated to build the feature vector for classification. The recognition performance of proposed method was evaluated on popular Japanese Female Facial Expression dataset using Support Vector Machine as the classifier. Extensive experimental results with prototype expressions show that proposed feature descriptor outperforms several popular existing appearance-based feature descriptors in terms of classification accuracy.