Abstract: The human face is very special. It is the crucial part of the body by which one recognizes a person visually. In this paper, a simple and an efficient method of emotion normalization for faces recognition is introduced. This approach is also used to reduce the computational overhead by 1/n (n is number of images for an individual in a training set) and minimize distance between source image and target images. This method is mainly based on deforming a source face to represent corresponding normal face using proposed Normal Face Diagram (NFD) in order to normalize the source face. An imaginary NFD based on the location of the eyes pair of the source face is drawn, the diagram consists of edges connecting vertex of the pair of eye brows, eyes, nose tip, end points of the lips and mid of the bottom lip. So source facial diagram (SFD) is drawn based on extracted feature points. The genetic algorithm is used to search for possible face region, while the Eigen face technique is used to determine the fitness of the region. After alignment of an eye point (in our method left eye is aligned) our method obtain the vectors of displacement between SFD and NFD then project the vertex of the SFD along with the corresponding vertex of the NFD by adjustment of the magnitude and direction of these motion vectors.