Efficient content-based retrieval of motion capture data

By Meinard Müller, Tido Röder, Michael Clausen. 2005

471 citations recorded by the source corpus.

The reuse of human motion capture data to create new, realistic motions by applying morphing and blending techniques has become an important issue in computer animation. This requires the identification and extraction of logically related motions scattered within some data set. Such content-based retrieval of motion capture data, which is the topic of this paper, constitutes a difficult and time-consuming problem due to significant spatio-temporal variations between logically related motions. In our approach, we introduce various kinds of qualitative features describing geometric relations between specified body points of a pose and show how these features induce a time segmentation of motion capture data streams. By incorporating spatio-temporal invariance into the geometric features and adaptive segments, we are able to adopt efficient indexing methods allowing for flexible and efficient content-based retrieval and browsing in huge motion capture databases. Furthermore, we obtain an efficient preprocessing method substantially accelerating the cost-intensive classical dynamic time warping techniques for the time alignment of logically similar motion data streams. We present experimental results on a test data set of more than one million frames, corresponding to 180 minutes of motion. The linearity of our indexing algorithms guarantees the scalability of our results to much larger data sets.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Control and Systems Engineering

In Signal Processing

Citations in the corpus

Fusing Geometric Features for Skeleton-Based Action Recognition Using Multilayer LSTM Networks (2018, unknown): “[27] introduce a class of Boolean features expressing geometric relations”
A Multimedia Retrieval Framework Based on Semi-Supervised Ranking and Relevance Feedback (2012, background): “In recent years, researchers also proposed numerous content-based retrieval systems for new media types, including 3D data, cultural artifacts, motion data, and biological data [4], [20], [22].”
EDGE: Editable Dance Generation From Music (2022, methodology): “Diversity Diversity metrics are computed following the methods of previous work [33, 53], which measure the distributional spread of generated dances in the “kinetic” (Dist k ) and “geometric” (Dist g ) feature spaces, as implemented by fairmotion [13, 39, 41].”
Documentation Mocap Database HDM05 (2007, unknown): “1.4.”
3D skeleton-based human action classification: A survey (2016, background): “In [108], only 31 boolean features are manually identified and used for content-based motion retrieval.”
Posebits for Monocular Human Pose Estimation (2014, methodology): “Pose attributes have been used for content-based MoCap retrieval [14].”
Pose-based human action recognition via sparse representation in dissimilarity space (2014, methodology): “Projections of velocity vectors calculated between consecutive poses, on a plane defined by the joints corresponding to shoulders and heaps have been used in [21].”
A Review on Video-Based Human Activity Recognition (2013, methodology): “Moreover, the coordinates of the body modeling can be further converted into lower-dimensional or more discriminative features, such as polar coordinate representation [71], Boolean features [72] and geometric relational features (GRF) [73,74], for effective recognition purpose.”
Manifold Warp Segmentation of Human Action (2018, background): “After that, these motion clips can be used in human motion synthesis [1], recognition, retrieval [2], [3], and prediction [43].”
On Geometric Features for Skeleton-Based Action Recognition Using Multilayer LSTM Networks (2017, unknown): “[18] introduce a class of Boolean features expressing geometric relations between certain body points of a pose.”
Bailando: 3D Dance Generation by Actor-Critic GPT with Choreographic Memory (2022, methodology): “In concrete, for the dance quality, we calculate the Fréchet Inception Distances (FID) [13] between the generated dance and all motion sequences (including training and test data) of the AIST++ dataset on kinetic features [33] (denoted as ‘k’) and geometric features [32] (denoted as ‘g’), which are ”
A Survey on Human Motion Analysis from Depth Data (2013, background): “[38] where skeleton motion is encoded by relational pose features [39], as shown in Figure 12.”
AI Choreographer: Music Conditioned 3D Dance Generation with AIST++ (2021, methodology): “As prior work used motionencoders that are not public, we measure FID with two welldesigned motion feature extractors [62, 63] implemented in fairmotion [29]: (1) a geometric feature extractor that produces a boolean vector zg ∈ R(33) expressing geometric relations between certain body points in the”
Does Human Action Recognition Benefit from Pose Estimation? (2011, methodology): “2 Pose-based features One of the biggest challenges of using posed-based features is that semantically similar motions may not necessarily be numerically similar [15, 22].”