Moving fast and slow: Analysis of representations and post-processing in speech-driven automatic gesture generation

TL;DR

This paper introduces a new speech-driven gesture generation framework that leverages representation learning, improving motion dynamics and naturalness, and highlights the significance of post-processing in gesture synthesis.

Contribution

The paper extends deep-learning methods for gesture generation by analyzing input/output representations and the impact of post-processing, demonstrating improved naturalness and motion quality.

Findings

Improved motion dynamics and speed matching in generated gestures.

User studies show increased perceived naturalness of gestures.

Post-processing techniques like smoothing enhance gesture quality.

Abstract

This paper presents a novel framework for speech-driven gesture production, applicable to virtual agents to enhance human-computer interaction. Specifically, we extend recent deep-learning-based, data-driven methods for speech-driven gesture generation by incorporating representation learning. Our model takes speech as input and produces gestures as output, in the form of a sequence of 3D coordinates. We provide an analysis of different representations for the input (speech) and the output (motion) of the network by both objective and subjective evaluations. We also analyse the importance of smoothing of the produced motion. Our results indicated that the proposed method improved on our baseline in terms of objective measures. For example, it better captured the motion dynamics and better matched the motion-speed distribution. Moreover, we performed user studies on two different datasets. The studies confirmed that our proposed method is perceived as more natural than the baseline, although the difference in the studies was eliminated by appropriate post-processing: hip-centering and smoothing. We conclude that it is important to take both motion representation and post-processing into account when designing an automatic gesture-production method.