Generic Content-Based Retrieval of Marker-Based Motion Capture Data

Abstract

In this work, we propose an original scheme for generic content-based retrieval of marker-based motion capture data. It works on motion capture data of arbitrary subject types and arbitrary marker attachment and labelling conventions. Specifically, we propose a novel motion signature to statistically describe both the high-level and the low-level morphological and kinematic characteristics of a motion capture sequence, and conduct the content-based retrieval by computing and ordering the motion signature distance between the query and every item in the database. The distance between two motion signatures is computed by a weighted sum of differences in separate features contained in them. For maximum retrieval performance, we propose a method to pre-learn an optimal set of weights for each type of motion in the database through biased discriminant analysis, and adaptively choose a good set of weights for any given query at the run time. Excellence of the proposed scheme is experimentally demonstrated on various data sets and performance metrics.

Control Model for Collaborative Manufacturing: An Integrated Opened Framework for Human-Robot Collaboration

“…Markerless optical MOCAP represents a recent advance, freeing the operator to perform his/her activities. This technology is performed by proper computer vision algorithms, which analyze and process the images recorded by cameras to recognize the human and his motions from the background static scene [22]. Moreover, marker-less optical MOCAP technology avoids any interference with the worker activity, which is typical of the MOCAP solutions which adopt cumbersome suits.…”

Visual Browse and Exploration in Motion Capture Data with Phylogenetic Tree of Context-Aware Poses

“…Motion retrieval. The paradigm of query by example (QBE) of content-based motion retrieval (CBMR) draws the most attention of academia in the field of relevant motion search [3][4][5][6][7][8]. Just as content-based image retrieval (CBIR) [17], feature extraction and similarity measure are two most important parts in CBMR.…”

“…The challenges of this task lay in the different magnitude of human skeletons, high dimension of 3D poses and spatio-temporal deformation of homogeneous motions, etc. Over the time, lots of approaches have been suggested to tackle these problems [3][4][5][6][7][8], the majority of which fall into the paradigm of query by example (QBE) of content-based motion retrieval (CBMR).…”

Retrieval of spatial–temporal motion topics from 3D skeleton data

“…Du et al [9] model a bidirectional recurrent neural network (RNN) unit of the joint trajectories for each of the five segmented body parts, and the lower-layer RNN units are fused hierarchically to examine higher-layer motion representations. Lv et al [25] measure the distance between motion signatures by marker-based geometry, which shows outstanding retrieval performance on various datasets. Wang et al [38] represent body parts by eigenvectors and match the individual and integrated similarities based on the body segments.…”