基于表情分析和视线追踪的用户反馈采集技术

Research on user feedback acquisition based on expression recognition and sight tracking technology

用户反馈,是指使用某一产品的用户对其产品所提出的有关于产品的情况反馈。用户反馈采集有利于公司优化其产品,为用户提供更好的服务。传统的用户反馈采集方法如跨站跟踪、Cookie跟踪或观察流量信息,仅反馈用户浏览行为的信息,而忽略了用户的潜在兴趣。基于表情分析和眼球视线追踪技术的用户反馈采集核心技术能够反映用户在网页页面浏览时,无意识状态下自然流露的潜在兴趣。研究采用基于类Haar特征的面部检测的Adaboost算法,及基于深度学习的面部情感识别技术,使人类面部情感识别的正确率可达90%。同时使用深度学习方法,在没有高精度且昂贵的仪器条件下,仅借助笔记本电脑前置摄像头实现视线追踪的效果。测试比较3种不同的深度学习的网络结构实现视线追踪的准确率,其中效果最佳的一种网络结构的准确率可达49.60%。

Customer feedback refers to the feedback of customers who use a product and then give back a comment about the product.Feedback analytics services use customer generated feedback data to measure customer experience and improve customer satisfaction.Feedback data is collected,then,key performance indicators and feedback metrics is turned into actionable information for website improvement.This paper studies the core technology of feedback analytics services based on facial expression recognition and eye tracking for Web page browsing.The research has built a website that has a function to recognize human expression.The accuracy of expression recognition can reach 90%.The research also explores three different deep learning network structures and tests their accuracy of eye tracking.Instead of the help of high precision and expensive instruments,the research obtains the face pictures through the front camera of the laptop computer and uses the depth learning method to analyze the eye tracking.The accuracy of the best network structure among the three can reach 49.60%.