基于边缘云框架的高效安全人脸表情识别

Efficient and secure facial expression recognition method based on edge cloud framework

针对物联网环境下数据量大且人脸表情识别率低的问题,提出基于边缘云框架的高效安全人脸表情识别方法。物联网设备通过多秘密共享技术获取用户信息,并分发到不同的边缘云。边缘云利用语谱图和局部二值模式的方法提取语音特征,采用差值中心对称局部二值模式获得图像特征,将特征送至核心云。基于栈式稀疏去噪自编码器融合语音和图像特征,实现人脸表情的识别,并在RML和eNTERFACE’05数据库上进行实验。实验结果表明,该方法的识别准确率明显高于对比方法,抵御网络攻击的能力较强。

In view of the problem of large amount of data and low rate of facial expression recognition in the Internet of things environment,an efficient and secure facial expression recognition method based on edge cloud framework was proposed.The Internet of things devices acquired user information through multi secret sharing technology and it was distributed to different edge clouds.The edge cloud extracted speech features using spectrogram and local binary mode,and image features were obtained using the difference centrosymmetric local binary mode,and the features were sent to the core cloud.Based on the trestle sparse denoising self-encoder,the speech and image features were fused to realize facial expression recognition,and the experiments were carried out on RML and eNTERFACE’05 database.The results show that the recognition accuracy of proposed method is significantly higher than that of the comparison method,and the ability to resist network attacks is strong.