Space-Time Pose Representation for 3D Human Action Recognition

By Maxime Devanne, Hazem Wannous, Stefano Berretti, Pietro Pala, Mohamed Daoudi, A. del Bimbo. 2013

72 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition, Human-Computer Interaction

How researchers see this paper

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

In Biomedical Engineering

In Computer Vision and Pattern Recognition

In Human-Computer Interaction

Citations in the corpus

Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, methodology): “[187] Riemannian Manifolds Conc Manif Hand X X X X Zhu et al.”
Skeleton-Based Action Recognition Using Spatio-Temporal LSTM Network with Trust Gates (2017, neutral): “9% Space Time Pose [83] Geometric 91.”
A novel hierarchical framework for human action recognition (2016, methodology): “Some other histogram-36 based features are also proposed, like histograms of 3D joints [13], space time pose [17], histogram of oriented displacements [15], points in a lie group [19], etc. Yang et al. [12] used a 3D skeleton-based method to achieve higher recog-40 nition rates than the depth…”
Gesture Modeling by Hanklet-Based Hidden Markov Model (2014, methodology): “In [27], each action is represented by spatio-temporal motion trajectories of the joints.”
Efficient Temporal Sequence Comparison and Classification Using Gram Matrix Embeddings on a Riemannian Manifold (2016, methodology): “Performance is compablack against the Hankelets subspace angles [21], Hankelet-based HMM [31], 3D joints [40], and Space-time pose [9] methods.”
Human-object interaction recognition by learning the distances between the object and the skeleton joints (2015, background): “There are several works [10][11][12][13][14] relyed on skeleton information and developed features based on depth images for human object interaction recognition.”
Action Recognition Using Rate-Invariant Analysis of Skeletal Shape Trajectories (2016, methodology): “Some of these datasets are also used in the experimental studies described later.”
Informative joints based human action recognition using skeleton contexts (2015, methodology): “We obtain an accuracy similar to the work in [16,25,29] (see Table 5).”
Spatiotemporal representation of 3D skeleton joints-based action recognition using modified spherical harmonics (2016, methodology): “Fusion spatiotemporal Zhu et al. (2013) 87.90 HO3DJ Xia et al. (2012) 90.92 Space-time pose Rep. Devanne et al. (2013) 91.5 SHs Al Alwani and Chahir (2015) 91.65 Proposed approach 93.0 The performance of the proposed method is also assessed based on the G3D-Action dataset.”
A Human Activity Recognition System Using Skeleton Data from RGBD Sensors (2016, background): “[23] proposed representing human actions by spatiotemporal motion trajectories in a 60-dimensional space, since they considered 20 joints, each of them with 3 coordinates.”
3D skeleton-based human action classification: A survey (2016, methodology): “Methods such as [105,113] attempt to consider the geometric relations between skeleton sequences rather than body parts in the same skeleton.”
Grassmannian Representation of Motion Depth for 3D Human Gesture and Action Recognition (2014, background): “92 Space-time Pose Representation [19] 91.”
WALKING WALKing walking: Action Recognition from Action Echoes (2017, methodology): “To demonstrate the efﬁcacy of our combined ESN and CNN, we compare it to each component separately: 1) ESN alone with mean-pooling and softmax; 2) CNN alone with the fusion strategy MSSC; 3) CNN alone with MSMC. Figure 6 shows that muli-step-fusion based ConvESN Zhu et al. , 2013] 87.90 LTBSVM [Slam”
Action recognition on motion capture data using a dynemes and forward differences representation (2014, methodology): “In [22], the authors propose a method for 3D human action recognition.”
3D-based Deep Convolutional Neural Network for action recognition with depth sequences (2016, methodology): “…One, random 1/3 of the Table 4: The performance of our method on the UTKinect-Action3D dataset, compared to the state-of-the-art approaches 90.92% Devanne et al. (2013)[51] 91.5% Chrungoo et al. (2014)[52] 91.96% Vemulapalli et al. (2014)[11] 97.08% 3D 2 CNN 82% JointVector 93% Proposed 96%…”