Vision-Based Classification of Social Gestures in Videochat Sessions

TL;DR

This paper presents a vision-based system for recognizing social gestures like handshakes and hugs in video chat, aiming to enable real-time mediated social touch, with initial promising accuracy but requiring further improvements.

Contribution

It introduces a novel vision-based approach for classifying social gestures in video chat environments, integrating with mediated social touch systems.

Findings

Gesture recognition accuracy varies across gestures

Initial system demonstrates feasibility for real-time applications

Further work needed to enhance practical deployment

Abstract

This paper describes the design and evaluation of the vision-based classification of social gestures, such as handshake, hug, high-five, etc. This is a component of the mediated social touch systems, which can be incorporated into ShareTable and SqueezeBands system to achieve automated gestures recognition and transmission of the touch between the users in real time. The results from our pilot study show the recognition accuracy of each gestures, and they indicate that significant future work is necessary to improve its practical feasibility in the mediated social touch applications.