Bio-inspired Dynamic 3D Discriminative Skeletal Features for Human Action Recognition

By Rizwan Ahmed Chaudhry, Ferda Ofli, Gregorij Kurillo, Ruzena Bajcsy, René Vidal. 2013

166 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

RGB-D-based action recognition datasets: A survey (2016, neutral): “Hierarchy of LDSs(28 joints used) [95] 2.”
Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, methodology): “olaidis [162] Dynemes Representation BoW Lowlv Dict X Vantigodi and Radhakrishnan [163] Meta-cognitive RBF Stat Lowlv Hand X X X X Ohn-Bar and Trivedi [164] HOG2 Conc Lowlv Hand X X X Chaudhry et al. [165] Shape from Neuroscience BoW Body Dict X X Oﬂi et al. [166] SMIJ Conc Lowlv Unsup X X X Miranda”
Learning Actionlet Ensemble for 3D Human Action Recognition (2014, background): “[32] uses linear dynamic systems to model the dynamic medial axis structures of human parts and proposes discriminative metrics for comparing sets of linear”
3D skeleton-based human action classification: A survey (2016, methodology): “We further note that such kind of evaluation may pose some challenges for works such as [102,99,103,100], which use a pyramidal approach and need to see the whole sequence for recognizing the action.”
Learning Clip Representations for Skeleton-Based 3D Action Recognition (2018, methodology): “[46] used a set of Linear Dynamical Systems (LDSs) to encode a hierarchy”
Semi-supervised Learning with Encoder-Decoder Recurrent Neural Networks: Experiments with Motion Capture Sequences (2015, methodology): “One of the most popular and effective methods to counter this problem is the use of Long-Short-Term Memory networks (LSTMs), as presented by Hochreiter & Schmidhuber (1997).”
Human Action Recognition by Representing 3D Skeletons as Points in a Lie Group (2014, background): “In [2], a human skeleton was hierarchically divided into smaller parts and each part was represented using certain bio-inspired shape features.”
Spatio-Temporal Graph Convolution for Skeleton Based Action Recognition (2018, methodology): “The statistic model (Xia, Chen, and Aggarwal 2012), shape model (Chaudhry et al. 2013), and geometric model (Vemulapalli, Arrate, and Chellappa 2014) are often used to characterize the motion trajectories.”
Vision-based human activity recognition: a survey (2020, background): “Joint angles, which assume the human body as a system of rigid connected segments and the movement as an evolution of their spatial configuration [7, 29, 128].”
Fusing Geometric Features for Skeleton-Based Action Recognition Using Multilayer LSTM Networks (2018, unknown): “As shown in Table IX, using JL d can achieve the 100% accuracy, the same as several previous works [5], [6], [9], [25].”
Rolling Rotations for Recognizing Human Actions from 3D Skeletal Data (2016, background): “On the other hand, part-based approaches consider human skeleton as a set of connected rigid segments, and represent it with features like joint angles [18, 19, 31], bio-inspired 3D features [6], individual part locations [38], relative 3D geometry between parts [33], etc.”
Simultaneous Feature and Body-Part Learning for real-time robot awareness of human behaviors (2017, background): “Chaudhry et al. [24] implemented bio-inspired mid-level features to represent human activities based on 3D skeleton data, by leveraging the ﬁndings in the research area of static shape encoding in the primate cortex’s neural pathway.”
A discussion on the validation tests employed to compare human action recognition methods using the MSR Action3D dataset (2014, methodology): “These results are improved in [70] with the same setup (83.89% for the whole dataset).”
Hierarchical recurrent neural network for skeleton based action recognition (2015, background): “[2], as well as DURNN-T, DBRNN-T, DURNN-L, DBRNN-L, HURNN-L.”
Efficient Temporal Sequence Comparison and Classification Using Gram Matrix Embeddings on a Riemannian Manifold (2016, unknown): “95.37 - - - - RBF Net [35] 97.58 - - - - Dynemes [19] 98.18 - - - - Bio-LDS [7] 100 - - - - HBRNN-L [11] Table 6: Classwise recognition accuracy (”