中文摘要：

针对图像Gabor变换计算代价和存储空间开销较高的问题，提出一种基于单演信号分析的人脸表情描述方法。该方法首先采用单演信号分析将人脸图像分解为单演幅度、相位和方向三个图像，并将其划分为多个矩形块子区域；然后在三幅图像的子区域上提取相应的由空间显著性加权的单演幅度、相位和方向二元模式特征直方图；最后将结合了空间显著性的三个加权特征进行融合增强特征的可分辨性。在JAFFE人脸表情数据库上的实验结果表明，该方法能有效提取人脸表情特征，提高人脸表情的识别率。与基于Gabor的特征相比，提出的方法具有更高的准确率和较低的特征维度。

英文摘要：

For the problem of high expenses including computational cost and storage space of image＇ s Gabor transformation, this paper proposed a novel facial expression description method based on monogenic signal analysis. The proposed method firstly adopted monogenic signal analysis to decompose a facial image into three monogenic amplitude, phase and orientation maps which then were divided into multiple rectangle sub-regions. Then, it respectively extracted monogenic amplitude, phase and orientation binary pattern feature histograms weighted by corresponding spatial saliency of sub-region on three maps. At last, it fused three weighted descriptors combined with spatial saliency to enhance discriminative power. Experimental results on JAFFE face expressional database demonstrate that the proposed approach can effectively extract facial expression feature and improve recognition rate. The proposed method has higher recognition accuracy and lower feature dimension compared to the Gabor-based features.