Dynamic emotion understanding using FCM based SVR in human-robot interaction

FCM based SVR is proposed for emotion understanding in human-robot interaction, where the real-time dynamic emotion recognition is realized by using Candide3 based feature point matching method, and emotional intention understanding is obtained mainly based on human emotions and identification information. It aims to make robots capable of recognizing and understanding human emotions, in such a way that make human-robot interaction run smoothly. Preliminary application experiments are performed in the developing emotional social robot system (ESRS), where six volunteers experience the scenario of 'drinking in the bar'. Results show that dynamic emotion recognition obtains 78.6% accuracy, and emotion understanding by using the proposed FCM based SVR model receives accuracy of 57.14% and 69.05% while C=2/3 (different genders/ages), which is 2.38% and 14.29% higher than that of SVR. Based on the preliminarily application experiments, the proposal is being extend to task mobile robot for behavior adaptation to customer's emotional intention in the developing ESRS.