基于CBP-TOP特征的人脸表情识别

Facial expression recognition based on CBP-TOP features

针对人脸表情时空域特征信息的有效提取,提出了一种CBP-TOP(centralized binary patterns from threeorthogonal panels)特征和SVM分类器相结合的人脸表情识别新方法。该方法首先将原始图像序列进行图像预处理,包括人脸检测、图像截取和图像尺度归一化,然后用CBP-TOP算子对图像序列进行分块提取特征,最后采用SVM分类器进行表情识别。实验结果表明,该方法能更有效地提取图像序列的运动特征和动态纹理信息,提高了表情识别的准确率。与VLBP(volume local binary pattern)特征相比,CBP-TOP特征在表情识别中具有更高的识别率和更快的识别速度。

According to effective extraction of facial expression information in space-time domain,this paper proposed a novel approach for facial expression recognition based on CBP-TOP features and SVM classifier.In this method,processed original image sequences first,including face detection,image interception and size normalized.Then extracted the features of image from the blocks of images using the CBP-TOP operator.Finally recognized six expressions by support vector machine classifier.The experiment result shows that,this method can extract movement feature of image sequences and dynamic texture information more effectively,as well as raise the accuracy of expression recognition.Compared with VLBP,CBP-TOP has greater improvement in recognition rate and recognition speed.