# Learning multimodal representations for sample-efficient recognition of human actions

## TL;DR

This paper introduces motion concepts, a multimodal representation of human actions that combines kinematics and context, and presents OMCL, an algorithm that learns and recognizes these concepts efficiently from minimal data.

## Contribution

The paper proposes motion concepts as a novel multimodal representation and introduces OMCL, a new algorithm for learning and recognizing human actions with high sample efficiency.

## Key findings

- OMCL outperforms standard algorithms in one-shot recognition tasks.
- Motion concepts effectively combine kinematic and contextual information.
- The approach demonstrates potential for sample-efficient human action recognition.

## Abstract

Humans interact in rich and diverse ways with the environment. However, the representation of such behavior by artificial agents is often limited. In this work we present \textit{motion concepts}, a novel multimodal representation of human actions in a household environment. A motion concept encompasses a probabilistic description of the kinematics of the action along with its contextual background, namely the location and the objects held during the performance. Furthermore, we present Online Motion Concept Learning (OMCL), a new algorithm which learns novel motion concepts from action demonstrations and recognizes previously learned motion concepts. The algorithm is evaluated on a virtual-reality household environment with the presence of a human avatar. OMCL outperforms standard motion recognition algorithms on an one-shot recognition task, attesting to its potential for sample-efficient recognition of human actions.