The skin of the human palm contains more than 20,000 tactile corpuscles. Depending on the skin depth, activation threshold, trigger mode, and other differences in tactile signal reception and cross-synergistic mechanisms, the skin can receive various types of tactile signals. By "calculating" the tactile signals in the nerve center of the brain, the object features can be perceived more comprehensively and specifically. Tactile visual reconstruction of perception is even possible, i.e., visualization of perception based on the structure, shape, texture, and other features of the tactile object. Inspired by this human tactile perception mechanism, tactile sensors that simulate the function of human skin have attracted great attention. So far, electrical sensors based on the principles of resistance, piezoelectricity, and triboelectricity can mimic the tactile nerve and collect and process physical information by monitoring the change in the sensor's electrical output signal during the contact process. However, electrical tactile sensors also have problems, such as potential leakage, slight corrosion, lack of resistance to electromagnetic interference, low sensitivity, and slow response speed. In contrast, using optical means as information carriers to realize tactile sensing has become an optional and ideal technological path, which has been demonstrated in multi-physical parameter fiber optic sensors.
To solve the above problems, the team led by Associate Researcher Yu Yang from the Micro-Nano Optoelectronics and Intelligent Perception Group of the National University of Defense Technology has proposed an optical micro-fiber array skin (OMAS) for object shape recognition in human-computer interaction. This OMAS uses a four-sided longitudinal and transverse micro-nano structure to successfully simulate the multi-functional tactile receptors of the human fingertip or subcutaneous layer and realize the synergistic effect of multi-tactile receptors in multiple tactile modalities (see Fig. 1(a)). To further realize human-like multimodal tactile visual reconstruction capability, the team integrated OMAS with the self-developed intelligent signal processing module and simulated the processing of bioelectrical signals by the human brain using machine learning algorithms such as the Fully-Connected Neural Network-FCNN, which realized the multi-functional perception and spatial reconstruction mechanism of object features such as shape, hardness, surface texture, etc. (see Fig. 1(b) and (c)).
Through experiments, the team showed that OMAS can be used as bionic flexible tactile skin for robots, i.e., acting as a multi-functional tactile receptor. As shown in Fig. 2, by analyzing static pressure data, OMAS can perceive the softness, hardness, and shape of contact objects very well (pressure recognition of six common objects with 100% accuracy). As shown in Fig. 3, by analyzing the characteristics of dynamic pressure tactile signals, OMAS can accurately identify the material and surface texture of contact objects (the recognition accuracy of ten fabrics is 98.5%, and the recognition success rate of ten digits 0-9 in internationally common Braille is 99%). As a proof of concept, the team integrated OMAS into a robot hand, successfully recognized Mahjong among multiple different objects, and realized the recognition of Mahjong suits and the reconstruction of perception. This verifies the advantage of this multiphotonic bionic skin in vectorial tactile perception, which is important for supporting the recognition of complex 3D textures on the surface of objects and even realizing tactile visualization and reconstruction perception.
https://www.oejournal.org//article/doi/10.29026/oea.2025.240152
