爪色センサによるヒトの触覚情報の触譜化とその分析

抄録

To record human hand techniques and shear them to humans and robots are our final goal. In this study, we develop a nail-colored tactile sensor that records human fingertip force and investigate tactile score as a method for recording time-series tactile data. Regarding the nail color tactile sensor, there was a problem in which the relative position of the nail and the sensor deteriorated each time the sensor was attached and detached, which deteriorated the estimation accuracy. We investigated the collection of training data and ways to improve estimation accuracy using the model. As a result of the investigation, we confirmed that the accuracy of vertical force estimation can be improved by attaching and detaching the sensor seven times when acquiring learning data and by using Xception. We also recorded vertical force using this sensor, converted the time series data into tactile score, and reproduced fingertip force. As a result, I found it a little difficult to reproduce the finger force while visually viewing the tactile score. In the future, we will verify the effectiveness of the method in transferring skills between humans and humans and between humans and robots.