基于面部检测和深度神经网络的面部情绪自动识别算法

面部情绪识别已成为可见光人脸识别应用的重要部 分,是光学模式识别研究中最重要的领域之一。为了进一步实现可见光条件下面部情绪的自 动识别,本文结合Viola-Jones、自适应直方图均衡(AHE)、离散小波变换(DWT)和深度卷 积神经网络(CNN),提出了一种面部情绪自动识别算法。该算法使用Viola-Jones定位脸 部和五官,使用自适应直方图均衡增强面部图像,使用DWT完成面部特征提取；最后,提取 的特征直接用于深度卷积神经网络训练,以实现面部情绪自动识别。仿真实验分别在CK+数 据库和可见光人脸图像中进行,在CK+数据集上收获了97%的平均准确 率,在可见光人脸图像测试中也获得了95%的平均准确率。实验结果 表明,针对不同的面部五官和情绪,本文算法能够对可见光面部特征进行准确定位,对可见 光图像信息进行均衡处理,对情绪类别进行自动识别,并且能够满足同框下多类面部情绪同 时识别的需求,有着较高的识别率和鲁棒性。

Automatic facial expression recognition algorithm based on facial detection and deep neural network

Facial emotion recognition has become an important part of visible face recognition and one of the most important fiel ds in optical pattern recognition.In order to further realize the automatic reco gnition of facial emotions under visible light,this paper proposes an automatic recognition algorithm of facial emotions combining viola-Jones adaptive histogr am equalization (AHE) and discrete wavelet transform (DWT) and deep convolutiona l neural network (CNN).The algorithm uses viola-jones to locate the face and fa cial features,uses adaptive histogram equalization to enhance the facial image,a nd uses DWT to complete facial feature extraction.Finally,the extracted features are directly used in deep convolutional neural network training to realize auto matic facial emotion recognition.The simulation experiment was conducted in CK+ database and visible face images respectively,and the average accuracy rate was 97% in CK+ data set and 95% in visible face image test.The experimental results show that the algorithm can accurately locate visible facial features according to different facial features and emotions.The algorithm presented in this paper can carry out balanced processing of visible image information and automatic rec ognition of emotion categories,and can meet the need of simultaneous recognition of multiple facial emotions under the same frame,with high recognition rate and robustness.