A Proposal for Emotion-Expressive Editor:EmoEditor by Font Changing
Yuki Shimamura, Michiharu Niimi · 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · 2022
Text media is one of important ways in communications on computers. For example, email, LINE or Twitter uses it frequently. In addition, captions in motion pictures are necessity for not only watching foreign cinema/TV but also one who is older or has hearing impairment. In general, a text character is recognized as rendered images on display. Therefore, it is difficult for text to express the emotions we have in writing sentences. In this paper, we propose a method to communicate emotions by changing font image. Firstly, we link a specific font and an emotion, then pair of a font data connected to an emotion and a standard font that is stored in computers is input to deep neural network (DNN) based on GAN, and we train it. An emotion is given to the neural network as a class label data. The input of the DNN is rendered font data. In other words, for the DNN, the input data is a standard font data we usually have with rendered, the output data is a generated rendered font image with emotion reflected. Furthermore, a mixed emotion font is able to be generated by adjusting parameters. We apply this neural network to editor system which is enable us to express emotions, which is called EmoEditor (Emotion-Expressive Editor). Through evaluation experiments, we confirmed the feasibility to communicate an emotion via an electronic media.