Construction of impression estimation models for the design of smartphone vibration feedback

Abstract

The importance of vibration feedback on mobile devices has been discussed in the human-computer interaction community. However, the psychological effects in the vibration feedback have not been much discussed. The goal of this study is to construct a design support system for vibration feedback, which considers user impressions. Toward constructing this system, a set of models to quantify the impressions is necessary as the first step. This paper describes the construction of the impression estimation models. We first quantified how the vibration feedback affects user impressions using the 85 various types of patterns. The result of the quantification revealed that users’ impressions can be controlled by changing the features of vibration patterns. On the basis of the results, we extracted three feature categories: intensity, suspension and change in intensity, and seven features in total were determined for these categories. Using the determined features and the quantification results, neural-network-based models were separately constructed for eight pairs of adjectives. The performance tests of the constructed models demonstrated sufficient accuracy. For all models, the average errors between the subjective evaluation and model output scores were within 7%. Also, all correlation coefficients between these scores showed a significant correlation.