Persistent-Transient Duality in Human Behavior Modeling

TL;DR

This paper introduces a neural network model that captures the persistent and transient aspects of human behavior, improving motion prediction by automatically learning the duality structure.

Contribution

It presents a novel parent-child multi-channel neural network with a transient switch to model human behavior duality, outperforming existing methods in motion prediction.

Findings

Superior performance in human-object interaction motion prediction

Automatic discovery of behavior duality structure

Effective management of transient sessions with a transient switch

Abstract

We propose to model the persistent-transient duality in human behavior using a parent-child multi-channel neural network, which features a parent persistent channel that manages the global dynamics and children transient channels that are initiated and terminated on-demand to handle detailed interactive actions. The short-lived transient sessions are managed by a proposed Transient Switch. The neural framework is trained to discover the structure of the duality automatically. Our model shows superior performances in human-object interaction motion prediction.