MOCAP signal interpolation using low-rank matrix recovery

We propose a new regularization method for 3D MOCAP (Motion Capture) signal restoration. As a MOCAP signal restoration method, methods considering low-rankness and smoothness or Truncated Nuclear Norm (TrNN) for low-rank approximation have been proposed. However, with these methods, the smoothness i...

Full description

Saved in:
Bibliographic Details
Main Authors: Imamura, Ryuji, Okuda, Masahiro
Format: Conference Proceeding
Language:English
Subjects:
Acceleration
Convex functions
Interpolation
Minimization
Optimization
Signal restoration
Smoothing methods
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We propose a new regularization method for 3D MOCAP (Motion Capture) signal restoration. As a MOCAP signal restoration method, methods considering low-rankness and smoothness or Truncated Nuclear Norm (TrNN) for low-rank approximation have been proposed. However, with these methods, the smoothness in the time direction of the MOCAP signal cannot be sufficiently restored. In this study, from the a priori knowledge that many of human movements involve smooth acceleration/deceleration and the a priori knowledge that the acceleration/deceleration patterns are similar among related joints, we introduce the regularization based on the second-order differentiation in the time direction. In numerical experiments, we apply this method to MOCAP signal restoration with continuous missing in the time direction and show that the performance is superior to the conventional methods.
ISSN:2640-0103
DOI:10.23919/APSIPA.2018.8659679