Abstract: As one of the more effective applications that used in image understanding and analysis, face recognition has given significant attention in last years. There are several methods used for face recognition such as PCA, LDA, Zernike and each method has limitations and strength points. This study offers a statistical estimate of the execution to recognized the human faces in digital images by using a new feature extraction method which based on hybrid system for face recognition that contains: Gabor filters and singular value decomposition. By using the Gabor filters 40 sub-images were obtained from the original images in 5 scales and 8 orientations and by SVD using one matrix U from 3 matrices [USV] that have singular value which represent feature extracted from an image and using hybrid technique normalize features vectors by Z-score to get optimal values then fusion two features vectors 2D Gabor filter and SVD to optimize the recognition rate. Finally, classification step is done by (Euclidean distance) to take the decision about matching. The outcomes experimental showed that the suggest system is effective, it has been tested using ORL face images databases with 10 cases and achieved recognition rate from 77.7-100%, also, applied on FEI Brazil face database with 5 cases and achieved recognition rate from to 84-100%.