Abstract
Quantitative relationship between Kansei and physical quantity is necessary to design Kansei Quality. In general, such relationship is obtained from sensory evaluation with participants. Number of samples that participants can evaluate is limited. Thus, efficient experimental design under limited number of samples is required. In this study, we consider prior estimation of experimenters and past experimental data. We adapted QFD matrix to obtain experimenter's prior estimation or intuition and combined design parameter using the matrix. We apply Bayesian Statistics to integrate the prior estimation. We used hierarchical Bayesian model to model individual variation of regression coefficients, which often occur in human Kansei. . We modified Kernel density estimation to construct confidence index for regression function in consideration of function sensitivity.
