Tactile Augmentation of Material Classification via Imperceptible On-Skin Triboelectricity Collection.

Harnessing intrinsic triboelectric signals from human skin holds promise for enhancing tactile perception. However, collecting these signals without disrupting normal skin functions and convoluting motion artifacts remains challenging. Additionally, person-to-person signal variance complicates data processing. In this study, it is demonstrated that triboelectric signals generated from touch can be imperceptibly collected and processed using a machine learning model to achieve tactile augmentation. When one hand contacts and rubs against a target object, charge transfer occurs between the skin and the object's surface. By placing a substrate-less microfiber electrode on the finger of the other hand, a body-coupled triboelectric circuit is formed to collect these signals, which contain material-specific features such as amplitude and peak ratio. A machine learning technique is developed to process the triboelectric signals, enabling the classification of six different materials with a prediction accuracy of ≈95%. The material differentiation model is further validated across different users, achieving an overall accuracy of ≈88 %, illustrating the potential of utilizing the body-coupled triboelectric circuit for tactile augmentation.