基于离散Shearlet类别可分性测度的人脸表情识别方法

Facial expression recognition based on separability assessment of discrete Shearlet transform

针对优化表情特征稀疏表达问题,提出一种基于离散Shearlet类别可分性测度的人脸表情识别方法。首先,对预处理后的图像进行离散Shearlet变换,得到变换系数。然后,根据测度函数评估每个方向与尺度系数的可分性指标,在最佳可分性方向与尺度上,融合低频和高频系数作为特征。最后,引入支持向量机进行分类。结果证明:本文方法选取具有最佳可分性的尺度和方向系数作为特征,抛弃了无用信息,降低了特征维度与计算量,使系统更高效。

To solve the problem of sparse expression of facial expression features,a facial expression recognition method based on separability assessment of the Discrete Shearlet Transform(DST) is proposed. DST is a relatively new image multiscale geometric analysis method. First,the DST transform is applied to the preprocessed facial expression images,and the transformation coefficients are obtained.Then,according to the separability evaluation function,the separability index in each direction and the scale coefficients are evaluated, and the low and high-frequency coefficients are fused on the best separability direction and scale. Finally,Support Vector Machine(SVM)is introduced to classify the facial expression. The experimental results show that the proposed method can select the best separability scale and direction coefficient as the feature,abandon the useless information,reduce the feature dimension and computation cost,therefore,the system is more efficient.