Identifying gender from human faces using correlation filters

Facial gender recognition plays an important role in various industrial applications such as human-computer interaction and targeted advertising. Although several methods have been applied to facial gender recognition, it is still considered as a challenging problem. In this paper, a system based on optimal trade-off (OT) - Maximum average correlation height (MACH) filter is developed for facial gender recognition. OT-MACH filter is a special method in the domain of correlation filters. Correlation filters have shown promising performance results in areas related to object recognition. Correlation filters are attractive due to their noise tolerance and shift invariance properties. Extensive experiments are performed to assess the capability of the OT-MACH filter for gender identification using FERET dataset. The system achieves an error rate of 3.5% on this dataset.