Automatic Estimation of Multidimensional Ratings from a Single Sound-Symbolic Word and Word-Based Visualization of Tactile Perceptual Space.

Abstract

Several pairs of Japanese adjective words pertaining to material's properties, such as roughness and hardness, have been used in Japanese studies to quantitatively evaluate variations in tactile sensations. This method asks observers to analyze their perceptual experiences one by one. An alternative notion is that human perceptual recognition is performed as a whole rather than by using fragmented factors. Based on this notion, we propose a system that can automatically estimate multidimensional ratings of touch from a single sound-symbolic word that has been spontaneously and intuitively expressed by a user. When a user inputs a sound-symbolic word into the system, the system refers to a database of phonemes and their auditory impressions, and calculates ratings in terms of 26 pairs of fundamental scales of touch. The estimated ratings of sound-symbolic words enable us to visualize a tactile perceptual space. Our study provides an alternative method for estimating the fine quality of tactile sensations.