Experiments on Stochastic Resonance Toward Human Mimetic Tactile Data Processing

Abstract

In the present research, human tactile stochastic resonance (SR) capable of enhancing sensitivity by superimposing proper noise upon undetectable weak signals is utilized to enhance the tactile processing method for social robotics. We develop an experimental apparatus composed of a piezoelectric actuator and its controller, and generate a step several microns high mixed with noise to perform a series of psychophysical experiments. Since psychophysical experiments are conducted based on the Parameter Estimation by Sequential Testing (PEST) method, we produce a PEST program that generates a stimuli sequence based on PEST. The experimental result shows that variation in the difference threshold (Difference Limen; DL) has a local minimum point in the relationship between DL and noise. Therefore, the tactile sensation’s just noticeable difference (JND) is decreased by appropriate external noise. Since JND denotes the scale divisions of sensation in the human mind, the present result shows that precise tactile sensations are enhanced by the appropriate external noise. Finally, we introduce a neural network model composed of nonlinear neurons with the bi-stable equilibrium condition to explain this result. Although original sensor data do not represent the morphology of the fine texture, the neural network model extracts the morphology and distinguishes the wave amplitude of the fine texture.