Aiming at the One-sidedness of face features or hair features to identify gender in the past, a novel method based on fusion of these features for gender classification was proposed. Face internal features were extracted by Gabor wavelet transform which are robust to the illumination change and scale variations, and feature dimensions were reduced by the method of PCA. Hair region was obtained by the method of dynamic searching in the area of face image. Two kinds of external features of hair length and hair surface area were defined and puted forward the method of corresponding feature extraction. To achieve nonlinear fusion of the three types of features with fuzzy neural network(FNN), the gender classification was completed in the Essex face database and a correct identification rate of 97.1% was obtained.