Low-latency compression of mocap data using learned spatial decorrelation transform

TL;DR

This paper introduces two low-latency mocap data compression frameworks utilizing a novel learned orthogonal transform (LOT) for spatial decorrelation, achieving higher compression efficiency with reduced computational cost and latency.

Contribution

The paper proposes a new learned orthogonal transform (LOT) for spatial decorrelation in mocap data and two frameworks that balance latency and compression performance.

Findings

Higher compression performance than state-of-the-art methods

Lower computational cost and latency

Effective spatial and temporal decorrelation

Abstract

Due to the growing needs of human motion capture (mocap) in movie, video games, sports, etc., it is highly desired to compress mocap data for efficient storage and transmission. This paper presents two efficient frameworks for compressing human mocap data with low latency. The first framework processes the data in a frame-by-frame manner so that it is ideal for mocap data streaming and time critical applications. The second one is clip-based and provides a flexible tradeoff between latency and compression performance. Since mocap data exhibits some unique spatial characteristics, we propose a very effective transform, namely learned orthogonal transform (LOT), for reducing the spatial redundancy. The LOT problem is formulated as minimizing square error regularized by orthogonality and sparsity and solved via alternating iteration. We also adopt a predictive coding and temporal DCT for temporal decorrelation in the frame- and clip-based frameworks, respectively. Experimental results show that the proposed frameworks can produce higher compression performance at lower computational cost and latency than the state-of-the-art methods.