Perceptual Coding of High-Resolution Fingertip Tactile Feedback

Abstract:
This paper presents a perceptual tactile codec designed for high-resolution tactile feedback at the fingertip. Unlike previous methods that primarily target single-point vibrotactile compression, our approach addresses the efficient transmission of high-resolution, multi-point tactile signals. We first preprocess the tactile signal to extract force field information, including normal and two orthogonal shear forces. By mapping the three channels of force data to a YUV format video stream, the data becomes compatible with standard video processing pipelines. To improve perceptual realism, we introduce a ROI-aware adaptive quantization method that considers the varying sensitivity across different regions of the fingertip. More bits are allocated to areas that are perceptually more important. We evaluate the performance of our method using ROI-PSNR and perceptual quality metrics, including weighted ROI-PSNR and HPW-PSNR. The results demonstrate that our approach outperforms the baseline HEVC coding scheme across all evaluation criteria.