Making Tactile Textures with Predefined Affective Properties

Abstract

Abstract-A process for the design and manufacture of 3D tactile textures with predefined affective properties was developed. Twenty four tactile textures were manufactured. Texture measures from the domain of machine vision were used to characterize the digital representations of the tactile textures. To obtain affective ratings, the textures were touched, unseen, by 107 participants who scored them against natural, warm, elegant, rough, simple, and like, on a semantic differential scale. The texture measures were correlated with the participants' affective ratings using a novel feature subset evaluation method and a partial least squares genetic algorithm. Six measures were identified that are significantly correlated with human responses and are unlikely to have occurred by chance. Regression equations were used to select forty eight new tactile textures that had been synthesized using mixing algorithms and which were likely to score highly against the six adjectives when touched by participants. The new textures were manufactured and rated by participants. It was found that the regression equations gave excellent predictive ability. The principal contribution of the work is the demonstration of a process, using machine vision methods and rapid prototyping, which can be used to make new tactile textures with predefined affective properties.

Layered-Modeling of Affective and Sensory Experiences using Structural Equation Modeling: Touch Experiences of Plastic Surfaces as an Example

“…It is valuable to refer to earlier studies where physical properties and affective or perceptual responses were linked (e.g. [1], [18], [19]) to develop multilayered models of products.…”

Varying Spread Fuzzy Regression for Affective Quality Estimation

“…To show the superior of the proposed method, we consider the relative improvements between the fuzz criteria achieved by the proposed methods to that achieved by the existing method. The relative improvement is formulated in (17). It is similar to the variation measure which indicates performance difference between two models [32].…”

“…These prediction models are generally developed by the statistical regression method as they provide explicit information such as variable significances, variable interactions and confidence intervals for dependent variables [7]. Recent research also shows that statistical regression models have been developed for the affective design of websites [5], aesthetic dental restorative materials [8], aesthetic plastic surgery [9], aesthetic evaluation of structural landscaping [10], visual aesthetics of images [11,12] or video [13], and the affective design of new products [14][15][16] or tactile textures [17].…”

Could linear model bridge the gap between low-level statistical features and aesthetic emotions of visual textures?

“…Bundgaard's work addressed the phenomenology of aesthetic experience, which showed why and how aesthetic experience should be defined relative to its object and the tools for meaning-making specific to that object, and not relative to the feeling [45]. Elkharraz, Thumfart, Akay, et al designed and manufactured 3D tactile textures with predefined affective properties, synthesized 48 new tactile texture using mixing algorithms and that were likely to score highly against the predefined affective properties [46]. Above all, visual scientists will come to realize that a rich resource they are provided with by artists who exploit the potential of the visual brain in their creations that are well worth studying scientifically [47].…”