Recognizing affection for a touch-based interaction with a humanoid robot

In order to facilitate integration into domestic and public environments, companion robots can seek to communicate in a familiar, “socially intelligent” manner, recognizing typical behaviors which people direct toward them. One important type of behavior to recognize is the displaying and seeking of affection, which is fundamentally associated with the modality of touch. This paper identifies how people communicate affection through touching a humanoid robot appearance, and reports on the development of a recognition system exploring the modalities of touch and vision. Results of evaluation indicate the proposed system can recognize people's affectionate behavior in the designated context.