基于智能感知的人脸细微表情情绪推定算法

Emotion Estimation Algorithm of Facial Subtle Expressions Based on Intelligence

人脸局部纹理的细微变化是判定情绪的关键视觉线索,但是表情的细微变化捕捉难度较大,且特征提取易出现偏差,提出一种基于智能感知的人脸细微表情情绪推定算法。通过滤波去除人脸图像干扰,采用差分法提取出连续图像中的人脸区域。使用Fisher线性变换确定RGB向量的最佳投影方向,区分人脸局部特征区域,提取细微表情特征。加入变分自编码器,将注意力模型嵌入到变分自编码网络中,根据网络关注推定不同类型情绪的重要特征。将未标记的特征输入到变分编码器中展开训练,同时将提取到的特征作为特征提取器,智能感知不同类型的面部表情,最终实现人脸细微表情情绪推定。仿真测试结果表明,所研究算法能够精准提取人脸细微表情特征,通过细微表情推定的情绪精度也较高,查全率和查准率可达97%,验证了该算法的可应用性。

The subtle change of local face texture is the key visual cue to judge emotion.However,it is difficult to capture the subtle change of expressions.Therefore,an emotion estimation algorithm of facial subtle expressions based on intelligent perception was proposed.At first,the interference on the face image was removed by filtering.Then,the face region in the continuous image was extracted by the difference method.Moreover,Fisher linear transformation was used to determine the best projection direction of the RGB vector and distinguish local feature areas of the face,and thus to extract subtle expression features.Meanwhile,the variational auto-coder was introduced,and then the attention model was embedded into the variational auto-coder network.According to the network attention,the important characteristics of different emotions were inferred.After that,the unlabeled features were input into the variational encoder for training.At the same time,the extracted features were used as a feature extractor to intelligently perceive different types of facial expressions.Finally,the emotion estimation of facial subtle expressions was achieved.Simulation results show that the proposed algorithm can accurately extract the features of facial subtle expressions.The emotional accuracy estimated by subtle expression is high as well.In addition,the recall and precision can reach 97%.