Single View Human Action Recognition using Key Pose Matching and Viterbi Path Searching

By Fengjun Lv, Ramkant Nevatia. 2007

486 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

HOPC: Histogram of Oriented Principal Components of 3D Pointclouds for Action Recognition (2014, methodology): “Some methods infer the 3D scene structure and use geometric transformations to achieve view invariance [5,9,15,26,39].”
Every Moment Counts: Dense Detailed Labeling of Actions in Complex Videos (2015, background): “Lv and Nevatia [16] represented actions as a sequence of synthetic 2D human poses rendered from different view points.”
Learning a non-linear knowledge transfer model for cross-view action recognition (2015, methodology): “Several techniques have been proposed for discriminative action representation such as 2D shape matching [20, 22, 37], spatio-temporal in terest points [5, 16, 27, 35], and trajectory-based representa tion [30-32, 36].”
View-Invariant Action Recognition from Point Triplets (2009, unknown): “Ç”
Video-Based Abnormal Human Behavior Recognition—A Review (2012, methodology): “1) Supervised: These methods build models of normal or/and abnormal behavior based on the labeled data [21], [102], [119], [120].”
STOP: Space-Time Occupancy Patterns for 3D Action Recognition from Depth Map Sequences (2012, methodology): “The methods in the ﬁrst category use global features such as silhou-ettes [10,8] and space-time volume information [15,4].”
Learning human actions via information maximization (2008, result): “Though it is difficult to directly compare our approach with [20] and [24], we obtained competitive performance, noting that our approach does not require 3D model construction.”
Does Human Action Recognition Benefit from Pose Estimation? (2011, background): “An alternative line of work models the entire body as a single entity, using silhouettes or visual hulls [2, 3, 20, 29, 30].”
Cross-view action recognition via view knowledge transfer (2011, background): “Some leading representations include space-time pattern templates [3, 27], shape features [36, 24, 7, 19], interest point based representations [15, 26, 29, 20], learned geometrical models of the human body parts [12], and motion/optical flow patterns [1, 25, 36].”
View-Invariant Action Recognition Based on Artificial Neural Networks (2012, methodology): “Method [38] Method [39] Method [40] Proposed method”
A survey on vision-based human action recognition (2010, methodology): “Related work by Lv and Nevatia [82] uses an Action Net,”
Multi-view 3D Human Pose Estimation in Complex Environment (2011, background): “M. Hofmann · D.M. Gavrila ( ) Intelligent Autonomous Systems Group, Informatics Institute, University of Amsterdam, Science Park 107, 1098 XG Amsterdam, The Netherlands e-mail: d.m.gavrila@uva.nl M. Hofmann e-mail: mhofmann.uva@gmail.com”
Structured Time Series Analysis for Human Action Segmentation and Recognition (2014, unknown): “For depth sensor data, xt is the vector representation of tracked joints.”
Fast Exact Hyper-graph Matching with Dynamic Programming for Spatio-temporal Data (2014, unknown): “Causality with respect to time has motivated sparse graphical models [7,36] in which inference can be carried out by means of dynamic programming.”
Expandable Data-Driven Graphical Modeling of Human Actions Based on Salient Postures (2008, methodology): “Considering the temporal nature of the human motion, we measure the similarity between two poses in terms of joint shape and motion, rather than shape or motion alone as used in most extant work [13], [14].”
Exploring the Trade-off Between Accuracy and Observational Latency in Action Recognition (2012, methodology): “Lv and Nevatia (2007) represent actions using a series of 2D human poses and perform silhouette matching between input and key frames.”
Recognizing human actions by attributes (2011, background): “More recently, great progress has been made by introducing more descriptive action representations such as space-time pattern templates[1, 21], 2D shape matching [33, 7], optical flow patterns [6, 33], trajectory-based representation [27], and bag-of-video-words [24, 19, 20].”