Fig. 1. Design and fabrication of a multiphoton neuron tactile skin. (a) The design concept and spatial reconstruction workflow of the multiphotonic neuron haptic skin for simulating the tactile perception and spatial reconstruction process of human subcutaneous multitactile cell fusion. (b) Flowchart of the interaction of each module in the spatial reconstruction process of mahjong by multiphotonic neuron haptic skin. (c) The structure of the multiphotonic neuron tactile skin, which consists of three layers: a silicone contact layer, an OM array embedded PDMS sensing layer, and a glass substrate layer.

Fig. 2. Validation of shape and hardness recognition ability of multiphotonic tactile neurons. (a) Schematic of the force applied to a single sensing unit of a multiphotonic neuron. (b) Changes in the path of light passing through the waist region under three forces. (c) Stress and deformation diagrams (indicated by color bars) of the waist region of a single sensing unit under three forces. (d) Optical power response plots of a single sensing unit under 0 to 3 N normal contact force in steps of 0.2 N, with each force transformation held for 5 s. (e) Normal force sensitivity from 0 to 3 N, with error bars indicating slight variations in optical power from the response time. (f) Individual sensing unit optical power response when pressure (1 N) is repeatedly applied more than 5000 times. The inset shows the optical power response in one of the two time domains after zooming in. (g) Pressure recognition based on object hardness and shape for six species using FCNN machine learning algorithm. (h) Visualization of clustered data clusters. (i) Confusion matrix of the six measured objects with 100% recognition accuracy.

Fig. 3. Validation of the material and surface texture recognition ability of micronized multiphoton neurons. (a) Neural network structure used for fabric recognition experiments with some signal acquisition results. (b) Accuracy change graphs of the training set and test set during the neural network training process, with the final accuracy stabilized at 99% for the training set and 98.5% for the test set. (c) Confusion matrix of ten fabric classification results in the validation set with 98.5% accuracy. (d) OMAS successfully realizes model training and online recognition of fabrics with the assistance of artificial intelligence algorithms. (e) Schematic diagram of 0-9 digit Braille. (f) Experimental process of real-time signal feedback and recognition of Braille phone numbers. (g) Real-time signal acquisition and recognition of Braille phone numbers displayed on the user graphical interface.

Fig. 4. Recognition of mahjong and its suits based on haptic parameters combined with spatial reconstruction of robotic arm. (a–c) With the help of AI algorithms, the multiphoton neuron haptic skin successfully recognizes different objects through the differences in hardness and shape, and the results are displayed on the PC screen. (d–f) Multi-photon neuron haptic skin successfully recognizes mahjong colors by differences in surface texture, and the results are displayed on the PC screen. (g, h) Computer interface for hardness and shape recognition and surface texture recognition.