Gender classification using spatial and temporal features

Automatic gender classification has immense applications in many commercial domains. In the paper, spatial and temporal feature based gender classification technique has been proposed. In the first step, texture based features in the spatial domain are extracted by dividing the training images into no. of blocks. Covariance matrix and singular value decomposition method has been applied on each block to extract the features. Discrete Wavelet Transform (DWT) has been introduced in the second step to extract temporal features. The feature vectors of test images are obtained and classified as male or female by Weka tool using 10 fold cross validation technique. The proposed approach provides 98% recognition rate on GTAV database while 91% on FERET database.