CNN-Based Electromagnetic Tomographic Approach for Simultaneous Tactile Imaging of Pressure and Temperature

Abstract

This paper introduces a novel electromagnetic tomographic approach for simultaneously imaging contact pressure and temperature using a single sensing material. The proposed sensor features adjustable detection ranges, along with a concise, scalable, and easily fabricated structure. Multi-frequency excitation elicits distinct voltage responses from pressure-induced displacement and temperature-induced conductivity changes, allowing decoupling based on their frequency-dependent patterns. These voltage features are processed by a convolutional neural network to reconstruct pressure and temperature distributions. The model developed using data with six excitation frequencies achieves good reconstruction performance on simulated data. Real-world experiments demonstrate the capability of the approach to coarsely reconstruct square-shaped pressure and temperature distributions, with noticeable residual modality coupling and discrepancies in intensity remaining. These results indicate the feasibility of the proposed approach and suggest its potential for multi-modal tactile imaging.