The human-like capabilities of robots are linked to their well-established perception systems. Tactile perception enables the robot to perceive the texture and grain of objects through touch. Robots equipped with tactile perception can grasp and manipulate objects more precisely and detect the characteristics and attributes of objects, which enhances their perceptual and cognitive abilities. Tactile perception provides robots with important advantages, helps them achieve human-like capabilities, and promotes the continuous development and innovation of robotic technology.

Herein, a bionic finger with a flexible shell, nail, finger bone, liquid, pressure-sensitive element, and temperature-sensitive element was developed based on the principle of liquid pressure conduction. The hardness, temperature, and texture sensing ability of the bionic finger were investigated; the tactile feature parameters of the bionic finger touching the textured surface were extracted; and classification and recognition of the fabric surface texture were achieved by the bionic finger using the support vector machine algorithm.

Results revealed that when the bionic finger applied pressure to three different materials, the rates of change in the pressure curve were in descending order: L_wo>L_fo>L_sp. These results were consistent with the hardness of the materials tested. The steepness of the temperature change curves obtained by the bionic finger touching the three materials was in descending order: T_ss>T_pb>T_wo, which aligned with the thermal conductivities of the materials. As the roughness of the fabric surface increased, the peak average value and average power increased. Thus, a positive correlation existed between the peak average and average power values and roughness, namely the higher the peak average and average power, the higher the roughness of the fabric. With increasing fineness of the fabric surface, the dominant frequency and the spectral centroid increased, resulting in an enhanced sense of fineness. A significant positive correlation existed between the sense of fineness and both the dominant frequency and the spectral centroid. The larger the dominant frequency and the spectral centroid, the higher the sense of fabric fineness. The average accuracy of fabric surface texture recognition using the bionic finger and support vector machine method, based on the peak average, average power, dominant frequency, spectral centroid, and six frequency band feature intensities, was 92.8%, which was higher than the average human subjective recognition accuracy of 88.8%.

The rate of change of the touch pressure curve and the temperature curve of the bionic finger can indicate the softness and thermal conductivity of an object, indicating that the bionic finger has the ability to perceive hardness and temperature. The peak average and average power of the bionic finger extracted from the touch vibration signal can characterize fabric roughness, while the dominant frequency and the spectral centroid can characterize fabric fineness, indicating the ability of the bionic finger to perceive roughness and fineness. The average recognition accuracy of the bionic finger is higher than that of human subjective recognition, indicating the efficient and superior capability of the bionic finger to recognize and classify textile surface textures compared to human judgment.

Although tactile perception plays a crucial role in human perception of the external world, human understanding of tactile perception remains limited due to the complexity of its mechanism and the multitude of perceptual units involved. The friction between surface textures and finger skin provides vibratory stimuli on the skin surface during tactile perception, thereby activating the somatosensory areas. It is necessary to evaluate the tactile perception of fine textures based on the friction behavior of skin and the related cortical activity in response to the texture stimuli.

Different fine stripe texture depths (5, 10, 15, 20, 25, and 30 μm) were designed and processed using laser engraving. The depth recognition threshold of tactile perception for fine texture was systematically investigated using subjective evaluation, surface friction and vibration, and the neurophysiological response of the brain. The effects of the texture stimulus intensity and neuronal excitability on tactile perception were verified by a single-channel neural mass model.

An increase in the fine texture depth was associated with an increase in the subjective human texture sense, the degree of correct texture recognition, and the proportion of deformation friction. The average depth recognition threshold of tactile perception was found to be 11.60 μm. The load index, the maximum spectral amplitude of the vibration signal, the recurrence parameter entropy, the length of the longest vertical line segment, and the peak of P300 exhibited a substantial positive correlation with the fine texture depth. The latency of P300 showed a substantial negative correlation with the fine texture depth. When the texture depth exceeded the depth recognition threshold of tactile perception, the maximum spectral amplitude and nonlinear characteristic parameters of the touch vibration signal increased remarkably. The main frequency of the vibration signal also increased to be within the perceptual frequency range of the Pacinian corpuscle. As a result, the vibration signal system transformed from a homogenous state to a disrupted state. Furthermore, the intensity and the area of activation of the brain regions, the neuronal activity of the brain, the processing intensity, and the tactile recognition speed of the brain increased remarkably. Amplitude of the main frequency of the simulated electroencephalogram (EEG) signal increased with an increase in the mean value of the input signal. This trend was consistent with that of the real EEG signal, which indicated that the increase in the tactile intensity due to the increase in the texture depth was one of the reasons for the increase in amplitude of the main frequency of the tactile EEG signal. The main frequency of the simulated EEG signal decreased with an increase in the ratio of excitatory synaptic gain to inhibitory synaptic gain. This trend was consistent with that of the real EEG signal, which indicated that the increased excitability of the neuronal populations excited by the increase in texture depth was one of the reasons for the decrease in the main frequency of the tactile EEG signal.

The depth recognition thresholds of tactile perception for fine stripe textures, the finger touch tribological behavior, the frequency domain and nonlinear features of the touch vibration signals, and the time and frequency domain features of EEG signals undergo remarkable variations during touching and sensing. The single-channel neural mass model can effectively simulate real EEG signals.