Adaptive Motion Data Representation with Repeated Motion Analysis

Abstract

In this paper, we present a representation method for motion capture data by exploiting the nearly repeated characteristics and spatiotemporal coherence in human motion. We extract similar motion clips of variable lengths or speeds across the database. Since the coding costs between these matched clips are small, we propose the repeated motion analysis to extract the referred and repeated clip pairs with maximum compression gains. For further utilization of motion coherence, we approximate the subspace-projected clip motions or residuals by interpolated functions with range-aware adaptive quantization. Our experiments demonstrate that the proposed feature-aware method is of high computational efficiency. Furthermore, it also provides substantial compression gains with comparable reconstruction and perceptual errors.

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

“…Recently, Hou et al [8] proposed the mocap data tailored transform (MDTT), which partitions the input motion into clips and then computes a set of data-dependent orthogonal bases by minimizing the least square of distortions. Computational results show that MDTT significantly outperforms the existing techniques (e.g., [11,6,23,5]) in terms of both compression performance and runtime. However, due to the overhead of explicitly storing the orthogonal bases, MDTT is less appealing for the short motion sequence.…”

“…Liu and McMillan [12] projected only the keyframes on the PCA bases and interpolated the other frames via spline functions. Motivated by the repeated characteristics of human motions, Lin et al [6] projected similar motion clips into PCA space and approximated them by interpolating functions with range-aware adaptive quantization. Observing that distortion to each of the joints causes a different overall distortion, Váša and Brunnett [7] proposed perception-driven error metric so that important joints have a higher precision than that of joints with small impact.…”

“…In this subsection, we compare our clip-based scheme with only two works, namely PCA-Rate Distortion Optimization (PCA-RDO) method [7], and the equal segmentation case of Mocap Data Tailored Transform (MDTT) method [8], which represent the state-of-the-art. See [7] and [8] for detailed performance evaluation on earlier works [11,6,5,4,23].…”

“…To date, many mocap compression algorithms have been proposed (see Section 2). Among these approaches, most are sequence-based (e.g., [3,4,5,6,7,8,9]) in that they process all the frames of a mocap sequence at a time. These methods are able to achieve high compression performance.…”