This paper proposes a facial expression recognition method based on a novel facial decomposition. First, seven regions of interest (ROI), representing the main components of face (left eyebrow, right eyebrow, left eye, right eye, between eyebrows, nose and mouth), are extracted using facial landmarks detected by IntraFace algorithm. Then, different local descriptors, such as LBP, CLBP, LTP and Dynamic LTP, are used to extract features. Finally, feature vector, representing face image, is fed into a multiclass support vector machine to achieve the recognition task. Experimental results on two public datasets show that the proposed method outperforms state of the art methods based on other facial decompositions.