An Expression Analysis Service based on Edge Cloud Computing

Abstract

Facial Expression Analysis (FEA) is widely applied in many areas such as marketing, health care and on-line learning. For considering performance and cost, FEA currently is implemented as a cloud service. However, uploading the streaming of face images from the end of networks to clouds will probably result in that the network bandwidth in Internet is exhaustively consumed, and finally the quality of the FEA service will degrade to an unacceptable level due to high network latency. To avoid this problem, this paper is aimed at proposing an expression analysis service system based on the architecture of edge cloud computing. The proposed system can efficiently provide users with the services of expression analysis while it need not consume a large amount of network bandwidth for data transmission from user clients to data centers. Moreover, it can provide a higher precision of expression analysis than related works by a method combining MTCNN, OpenFace and BPNN, and it can effectively select edge servers for reducing the time of expression analysis.