Automatic Generation of Dynamic Arousal Expression Based on Decaying Wave Synthesis for Robot Faces

The automatic generation of dynamic facial expressions to transmit the internal states of a robot, such as mood, is crucial for communication robots. In contrast, conventional methods rely on patchwork-like replaying of recorded motions, which makes it difficult to achieve adaptive smooth transitions of the facial expressions of internal states that easily fluctuate according to the internal and external circumstances of the robots. To achieve adaptive facial expressions in robots, designing and providing deep structures that dynamically generate facial movements based on the affective state of the robot is more effective than directly designing superficial facial movements. To address this issue, this paper proposes a method for automatically synthesizing complex but organized command sequences. The proposed system generated temporal control signals for each facial actuator as a linear combination of intermittently reactivating decaying waves. The forms of these waves were automatically tuned to express the internal state, such as the arousal level. We introduce a mathematical formulation of the system using arousal expression in a child-type android as an example, and demonstrate that the system can transmit different arousal levels without deteriorating human-like impressions. The experimental results support our hypothesis that appropriately tuned waveform facial movements can transmit different arousal state levels, and that such movements can be automatically generated as superimposed decaying waves.