Generative Adversarial Network を用いた触覚情報の合成において合成割合が生成結果に与える影響の検証

抄録

In recent years, many tactile displays have been developed. Haptic displays based on tactile information obtained from actual textures have high reproducibility. However, since these contents reproduce tactile sensations based on recorded information, it is necessary to record tactile information from actual textures in order to extend the contents. To solve this problem, we propose a method to generate uncollected tactile information from existing tactile information using Generative Adversarial Network (GAN). In this paper, we generate new tactile vibration information by learning 3-axis acceleration information obtained from actual textures using GAN and combining two existing tactile information. By creating a spectrogram of the generated haptic information, we visually compared and verified the changes in features caused by the ratio and combination of the haptic information.