Synthesizing fuzzy linguistic vocal responses by adapting perception of robot based on visual attention

This paper proposes a method for adapting robot's perception to produce fuzzy vocal responses about the size of an object based on the visual attention of the robot. In a human-human interaction, the humans may use vocal responses, which have qualitative terms such as "small", "large" etc. The actual quantitative meaning of those terms depends on spatial arrangement of the environment where the attention is focused on. Therefore, spatial information of the environment is analyzed to adapt robot's perception about the size of an object, which is in its vision field. A fuzzy logic based adaptive method has been introduced to build up relationship between qualitative terms and quantitative meaning. The proposed method is capable of adapting its perception to synthesize vocal responses that have uncertain qualitative terms. It has been implemented and tested on interactive robotic head (IRH). A survey has been conducted using human participants to identify actual human perceptions on test scenarios. Results of the developed system and the survey have been analyzed and outcome statistics are presented.