Research And Implementation Of Video Surveillance System For Agent Customer Service Supporting Emotion Recognition

With the increasing popularity of video customer service systems,when users can see the agent customer service staff,the emotional state of the agent customer service staff at work is particularly important.The traditional agent customer service monitoring system can only monitor the overall working hours of the agent customer service personnel,but cannot further understand the work quality status such as the service attitude of the individual agent customer service,and the work quality in the video customer service system scenario is mainly expressed by the agent customer service emotional state.This paper designs a system that can monitor the emotional state of each agent customer service staff at work,and infers the current emotional state by recognizing the facial expression of each agent customer service staff,thereby increasing work quality monitoring.The key task of this paper is below:(1)This paper firstly conducts in-depth research on the current facial expression recognition algorithms in various fields,and summarizes the two major deficiencies of the existing facial expression recognition algorithms:First,the existing facial expression recognition algorithms usually only work on image-based facial expression datasets.The construction of the model on the above causes the model to ignore the salient regions in the facial expression and the importance relationship between the frames in the expression video,which leads to a bottleneck in the emotion recognition ability of the model,which is difficult to further improve.Second,most of the current facial expression recognition models are constructed by limited experimental participants in the laboratory environment to make specific facial expressions,so it will be affected by identity features to a greater extent.In addition,The facial expression dataset is mainly based on American,and there are few facial expression datasets dominated by Asian cultures.This paper proposes a facial expression recognition model based on the spatiotemporal attention mechanism to break through the performance bottleneck of the existing facial expression recognition model,and designs an identity-aware layer to extract facial expression features and remove identity features.In the blank field of ethnic and cultural expression recognition,this paper integrates a large Asian facial expression dataset,and allows the model to model on the Asian facial expression dataset.Then,through a series of comparative experiments,the efficiency of the facial expression recognition model designed in this paper is verified.(2)This paper designs and implements an agent customer service video surveillance system that supports emotion recognition.This paper first divides the system into a data acquisition module,a facial expression recognition module,a persistence module and a visualization module.The facial expression recognition module adopts the spatiotemporal attention mechanism proposed in this paper.Facial expression recognition model.Finally,as a series of tests,it is verified that the system can meet the scene usage requirements and meet the design expectations.