Real-time Perception of Non-verbal Human Feedback in a Gaming Scenario

摘要

Social interactive robots require sophisticated perception and cognition abilities to behave and interact in a natural human-like way. The precise perception of behaviour of an interaction partner plays a crucial role in social robotics. Human feedback is one of the most important behavioural cues of human during conversation. Feedback is a signal that human sends to the interlocutor, consciously or unconsciously, that can convey four communicative functions. These functions: continuation, perception, understanding, and acceptance of the message, activate or inhibits different behaviours of human. Past systems focused mainly on the scripted scenarios irrespective of interaction partners feedback behaviour. This paper proposes a system for perception and interpretation of the interaction partner feedback from non-verbal cues. The proposed work has been tested in a 20 question gaming scenario in real time and it has shown promising results.