APPROACHES TO THE CLASSIFICATION OF COMMUNICATIVE GESTURES: THEORY AND APPLICATION TO COMPANION ROBOTS

The paper presents an approach to the classification of communicative gestures. We describe the gestures such that their signified can be expressed with the help of various signifiers. We classify gestures through the identification of communicative functions (the signified) and an invariant of the performance of the gesture (the signifier). A certain communicative function can be expressed by gestures with different sets of executive organs: with the help of facial expressions, movements of the head, hands or body. By distinguishing communicative functions, we solve the task of modeling complex communicative behavior of a companion robot. In the paper we explain the need of distinguishing communicative functions in the analysis of real human behavior in emotional situations (based on the multimodal corpus REC), and describe the contexts and invariants of communicative functions. Using the case of the feedback expectation communicative function we describe the “focus on the interlocutor’s response” invariant and different variations of the corresponding gestures: (1) eye focus on the addressee, (2) demonstration of one’s inaction while waiting for the addressee, (3) demonstration of the readiness to make a move, (4) nods in support of one’s own thesis or the interrogative movements which are requesting the interlocutor’s reaction. We also describe the robot’s control architecture which simulates the basic features of natural communication — communicative functions and gestures. The description of the invariant and gestures for a certain communicative function allows us to adjust the robot’s communicative responses to the incoming statement.