A Proposal of Video Evaluation Method Using Facial Expression for Video Recommendation System

Abstract

Recently, video sharing services and video on demand services like YouTube, Netflix and Amazon Prime Video have come to prominence worldwide. These services try to get the users’ evaluation of videos for the better recommendation. For example, some of them prepare the thumbs-up/down buttons and star buttons in their interface to get users’ explicit evaluation. However, the effect of this method is doubts. The reason is these method needs user’s conscious operation. Therefore, we need efficient method for getting users’ implicit evaluation of videos. In this study, we focus on facial expressions especially smile of users watching videos. In this paper, we have developed the system that can get user’s expression of users watching videos by some APIs, and investigated the relation between evaluation of videos by scoring and facial expression by using the proposed system. As a result of the experiment, positive correlations between the ranking by the user’s scoring and the ranking based on the amount of smile of users watching videos are shown in 22 of the 28 participants. In addition, strong positive correlations were found among 13 of them. This result suggests possibility of using the implicit evaluation like smiling as well as the explicit evaluation like a scoring. So, it is suggested that recommendation method based on facial expression is effective.