Abstract
Wave propagation is widely used in engineering and biology, because even small changes along a propagation path encode valuable information about the traversed medium. In this paper, we employ this concept in the context of vibrotactile object perception, a cornerstone problem in safe and effective human-robot interaction. We developed a robotic system that analyzes properties of grasped objects by injecting a controlled vibration through the grasped object, while measuring vibrations on both fingers and computing spectral residual changes. We investigate object-modulated responses in the frequency domain using spectral and statistical analysis techniques and build compact spectral and statistical descriptors. We validate the method across various material properties: elasticity (rigid to elastic: PLA 100% infill, TPU, FLEX, SortaClear 40, SortaClear 18, SortaClear 12), porosity (PLA infills of: 100%, 80%, 60%, 40%, 20%, hollow), and applied grip force (1 - 5 N). The results show consistent, property-dependent spectral signatures of elasticity, porosity, and grasp force across repeated probing. Finally, we demonstrate gentle-touch grasp analysis, enabling the tactile system to assist with efficient grasping in practical scenarios (e.g., strawberry ripeness tracking) and to provide invaluable tactile information on wave-propagation behavior across various environments. The proposed method has strong potential for precision agriculture and safe human-robot interaction.