Gender recognition is one of the most interesting problems in face processing. Gender recognition can be used as a preprocessing phase in many applications. In this work we compare different approaches for gender recognition task, in accuracy and generalizing. First, we use principle component analysis (PCA) and discrete cosine transformation (DCT), for feature extraction and dimension reduction. Additionally, we used Bayesian approach and support vector machine (SVM) too. Finally, we compare these approaches in accuracy and generalizing.