Abstract: In order to overcome the limitation of traditional Gabor features in expression recognition, an expression recognition method based on gradient Gabor histogram (GGH) features is proposed.Firstly the fusion rule of gradient direction is used for fusing the original Gabor features on the same scale and different directions after pretreatment.Then GGH features are extracted from block histogram features of Gabor feature fusion diagram.Finally, facial expression classification are achieved by using support vector machine.Experimental results in JAFFE database and Pain Expressions database show that GGH features are more real-time than the traditional Gabor features when the higher recognition rate is ensured.