Anticipation of Touch Gestures to Improve Robot Reaction Time

Nonverbal communication is a critical way for humans to relay information and can have many forms including hand gestures, touch, and facial expressions. Our work focuses on touch gestures. In typical systems the recognition process does not begin until after the communication has completed, which can create a delayed response from the robot. It may take time for the robot to plan the appropriate response to touch, which could delay the reaction time. We have trained an artificial neural network on features extracted from the Leap Motion Controller, and successfully performed early recognition of touch gestures with high accuracy.