VR-MDS: multidimensional scaling for classification tasks of virtual and real stimuli

Evaluating the perceptual similarity between virtual and real sensory stimuli has been a serious problem for virtual reality interface researchers for a long time. One of the most commonly used evaluation methods is a classification task where assessors classify randomly presented stimuli into multiple candidate types. The results of this method are summarized using two types of confusion matrices, which have different stimulus sets. The present study developed a method that computes the locations of simulated and real stimuli in a perceptual space on the basis of the two confusion matrices. The spatial distribution of the stimuli allows us to visually interpret the perceptual relationships between stimuli and their perceptual dimensionality. This method is recommended when the guidance index based on the answer ratios of the confusion matrices is fairly high.