Information regarding tactile sensation in friction signals with high uncertainty

An artificial tribosystem utilizing silicone-rubber sliding surfaces has been developed for simulating the friction phenomena that occur when human fingers are rubbed together in an attempt to acquire the tactile sensation of small particles. Although the friction signals acquired in the system have high uncertainty, two types of feature quantities, including the information about the particle size, were obtained—the elastic recovery in the normal signal as the slow information and the weighted mean frequency in the tangential signal as the fast information. The discrimination of particle size is realized by using one of the feature quantities to a certain extent, and a hybrid algorithm using both the feature quantities provides the discrimination with higher accuracy, which is comparable to or higher than the accuracy of human tactile systems. Consequently, the slow and fast information are complementary, and the relative reliability between the slow and fast information varies with a change in particle size.