A Multi-sensory Service System using Modular Dynamic Bayesian Networks and Utility Function

A multi-sensory service is to induce customer’s purchases through emotional stimuli. The system for this service needs to predict emotion and selects relevant stimuli depending on customer’s emotion. In this paper, we propose a stimuli selection system using modular dynamic Bayesian networks and utility function. The emotion prediction module uses dynamic Bayesian network to predict customer’s emotion, because the state of human emotion is changed depending on the current environment. The dynamic Bayesian network is modularized to reduce time complexity, because the system needs to aware the emotion quickly. The stimuli selection module decides the appropriate stimuli based on utility theory. The proposed system is verified by several experiments. The computational time of the proposed modular dynamic Bayesian networks is 0.032 sec. It is dramatically reduction in comparison of the inference time of the monolithic method. The accuracy of the system is verified through the obtained data. Key-Words: sensory service, emotion prediction, stimuli selection, modular dynamic Bayesian networks, utility function