Learning shift-invariant sparse representation of actions

TitleLearning shift-invariant sparse representation of actions
Publication TypeConference Papers
Year of Publication2010
AuthorsLi Y, Fermüller C, Aloimonos Y, Ji H
Conference Name2010 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Date Published2010/06/13/18
PublisherIEEE
ISBN Number978-1-4244-6984-0
Keywordsaction characterization, Action recognition, action retrieval, action synthesis, Character recognition, data compression, human motion capture, HUMANS, Image matching, Image motion analysis, image representation, Image sequences, Information retrieval, joint movements, large convex minimizations, learning (artificial intelligence), learning shift-invariant sparse representation, Matching pursuit algorithms, minimisation, Minimization methods, MoCap data compression, Motion analysis, motion capture analysis, motion disorder disease, motion sequences, orthogonal matching pursuit, Parkinson diagnosis, Parkinson's disease, Pursuit algorithms, shift-invariant basis functions, short basis functions, snippets, sparse linear combination, split Bregman algorithm, time series, time series data, Unsupervised learning, unsupervised learning algorithm
Abstract

A central problem in the analysis of motion capture (MoCap) data is how to decompose motion sequences into primitives. Ideally, a description in terms of primitives should facilitate the recognition, synthesis, and characterization of actions. We propose an unsupervised learning algorithm for automatically decomposing joint movements in human motion capture (MoCap) sequences into shift-invariant basis functions. Our formulation models the time series data of joint movements in actions as a sparse linear combination of short basis functions (snippets), which are executed (or “activated”) at different positions in time. Given a set of MoCap sequences of different actions, our algorithm finds the decomposition of MoCap sequences in terms of basis functions and their activations in time. Using the tools of L1 minimization, the procedure alternately solves two large convex minimizations: Given the basis functions, a variant of Orthogonal Matching Pursuit solves for the activations, and given the activations, the Split Bregman Algorithm solves for the basis functions. Experiments demonstrate the power of the decomposition in a number of applications, including action recognition, retrieval, MoCap data compression, and as a tool for classification in the diagnosis of Parkinson (a motion disorder disease).

DOI10.1109/CVPR.2010.5539977