基于ENM-Gabor差分权重的人脸表情特征提取方法

FACIAL EXPRESSION FEATURE EXTRACTION METHOD BASED ON ENM-GABOR DIFFERENTIAL WEIGHT

针对人脸子区域对表情识别分类的重要程度不同,提出一种基于Gabor小波特征和ENM(Eye,Nose,Mouth)差分权重的表情特征提取方法。通过对人脸眼睛、鼻子、嘴巴三个区域进行特征提取并自适应加以权重,有效地区分了不同区域对识别表情的重要程度。对预处理后的表情图像提取ENM区域Gabor特征;将表情图像与中性图像作差值计算得到ENM差分权重;将ENM-Gabor特征结合差分权重得到最终的表情特征并用BP神经网络进行分类。与其他方法在JAFFE表情库上进行对比实验,实验结果表明,该方法相比于传统Gabor特征提取有了明显的提高,且平均识别率达到99.3%。

Aiming at the different importance of facial expression recognition and classification in human face sub-region,this paper proposes a extraction method of facial features based on the Gabor wavelet sign and differential weighting of ENM(Eye,Nose and Mouth).It can effectively distinguish the importance of different regions to the recognition of facial expression by extracting the features from the eyes,nose and mouth on the human face and weighting them adaptively.Gabor features of the ENM region were extracted from the preprocessed facial expression images;the difference values of the facial expression images and neutral images were calculated to obtain the ENM differential weight;the ENM-Gabor feature was combined with the differential weight to obtain the final expression features,and classified by BP neural network.Finally,this method was compared with other methods in the JAFFE expression database.The experimental results show that it is significantly improved compared with traditional Gabor feature extraction,and the average recognition rate reaches 99.3%.