A new method called Null Space Diversity Fisher Discriminant Analysis (NSDFDA) was developed for face recognition. This method can extract important features from face images to help identify people accurately. By using a special optimization criterion, NSDFDA can find discriminant vectors in a way that is helpful for pattern analysis. The algorithm can handle small sample sizes well, which is important for many applications. Tests on the Yale database showed that NSDFDA is effective for recognizing faces.