Texture perception with a biomimetic optical tactile sensor

Abstract

In this study we assess the ability of the TacTip to classify a range of 12 textured stimuli with particle sizes ranging from 642 μ m to 15 μ m. We observe stick-slip events at the interface between sensor and stimulus which vary depending on the texture. We compare the use of marker position with marker velocity for encoding texture and found the velocity feature performed substantially better with a RMS error of ~ 1.4 stimulus classes compared with ~ 2.3 for the position feature. Marker velocity is analogous to rapidly adapting mechanoreceptors, such as Meissner’s corpuscles, which are thought to be used for roughness perception, therefore this result may be indicative of the biology of human touch.