Fig. 1. (Color online) Schematic illustration of the main fabrication process for the smart finger patch (SFP). (a) Manufacturing process of the microcilia-based inward bending resistive sensor, including spin-coating, magnetic field-guided microcilia self-assembly, and carbon black spray coating. (b) Fabrication of the magnetoelectric generator (MEG) based on magnetized micropillars, including template method and magnetization. (c) Integration and encapsulation of the SFP.

Fig. 2. (Color online) (a) Schematic diagram of the application of smart finger patch (SFP) in the field of human-computer interaction. (b) Three common phases of index finger sensing, contact-based drip, contact based command input and self-bending. (c) Schematic diagram of SFP corresponding to the three functional phases and (d) the partition mechanism.

Fig. 3. (Color online) (a) Resistive response of the microcilia-based bending sensor during the flexion process from −120° to 0° to 120°. (b) Resistance change rate when performing 50 cycles of inward bending at 5°, 10°, 20°, 40°, 60°, 80°, 90°, and 120° in the direction shown in the inset. (c) The response time and recovery time of the bending sensor. (d) Characteristic resistance response signals of two bending sensors equipped on the two joints of the index finger under four grasping states, indicating the bending conditions at the two joints.

Fig. 4. (Color online) (a) Schematic diagram and optical images of the MEG sensor based on micropillar arrays under normal pressure deformation and (b) schematic diagram of the relative magnetic flux change during the process. (c) Output voltage in the copper coil under pressure loading at different frequencies. (d) Output voltage in the copper coil under different indenter displacements. (e) Comparison of output voltage when sliding on smooth and the surface of artificial steps. (f) Comparison of output voltage when pressing on materials with different hardness.

Fig. 5. (Color online) Robotic tactile methods for bread identification using non-visual sensing techniques. (a) Signal recordings from the M1, R1, R2, and M2 sensors of the SFP-based system during a 5-s grasping cycle for five types of bread (donut, small soft bread, large soft bread, small hard bread, large hard bread). (b) Diagram of the deep learning network based on the transformer architecture. (c) Schematic representation of the robot's application of SFP tactile technology for the discrimination of bread types and quality assessment. (d) Graph of training and validation accuracy for the deep learning network. (e) Confusion matrix illustrating the classification outcomes of the deep learning network.