Learning Variable-Length Markov Models of Behavior

By Aphrodite Galata, Neil Johnson, David C. Hogg. Computer Vision and Image Understanding, 2001

311 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/1f1d18451b5ab0d1ba76e67c72bcf9c5a25f7e6f

How researchers see this paper

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

In Artificial Intelligence

In Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Layered representations for human activity recognition (2002, methodology): “More complex models, such as Parameterized-HMM (PHMM) [19], Entropic-HMM [1], Variable-length HMM (VHMM) [8] and Coupled-HMM (CHMM) [2], have been used to recognize more complex activities such as the interaction between two people.”
Trajectory-Based Anomalous Event Detection (2008, background): “[11], Brand and Kettnaker [12], Bashir et al.”
Layered representations for learning and inferring office activity from multiple sensory channels (2004, methodology): “Section 5 presents the architecture and implementation of a system named SEER that uses LHMMs, and describes the details of feature extraction, learning and classification used in the system.”
Action-Conditioned 3D Human Motion Synthesis with Transformer VAE (2021, methodology): “Statistical models have been employed in earlier studies [7, 17].”
Motion texture: a two-level statistical model for character motion synthesis (2002, methodology): “By discretizing state variables, a hidden Markov model (HMM) can be used to describe motion dynamics as well [6, 41, 14 ].”
Learning and detecting activities from movement trajectories using the hierarchical hidden Markov model (2005, methodology): “The hidden Markov model (HMM) and some of its extensions, such as the couple hidden Markov model (CHMM) or the variable length Markov model (VLMM), are efficient for representing and recognizing simple activities [5, 9, 15].”
Learning Motion Patterns of People for Compliant Robot Motion (2005, background): “[14, 48, 54]) and recognizing behaviors or plans (e.”
Expandable Data-Driven Graphical Modeling of Human Actions Based on Salient Postures (2008, methodology): “Researchers have so far explored various types of visual information to describe human motion, including motion trajectories [2]–[4], sequences of silhouettes or contours of the human body [5], [6], spatio–temporal salient points [7], hierarchical configuration of body parts [8], [9], such as torso,”
Interactive motion generation from examples (2002, methodology): “The resulting Markov chain can be searched using dynamic programming to ﬁnd a motion that connects two keyframes [Molina-Tanco and Hilton 2000] or used in a variable length Markov model to infer behaviors [Galata et al. 2001] or directly sampled from to create new motions [Bowden 2000].”
Recent developments in human motion analysis (2003, methodology): “In addition to three approaches described above, the Principle Component Analysis (PCA) method [142,143] and some variants from HMM and NN such as Coupled Hidden Markov Models (CHMM) [139], Variable-Length Markov Model (VLMM) [144] and Time-Delay Neural Network (TDNN) [145], have also appeared in th”
Modeling transition patterns between events for temporal human action segmentation and classification (2015, methodology): “Valstar et al. presented a hybrid SVM/HMM system to segment a facial action into temporal phases (e.g. onset, offset, peak, and neutral states), with a noticeable performance gain [23], [24].”
Qualitative and Quantitative Spatio-temporal Relations in Daily Living Activity Recognition (2014, methodology): “A hierarchical approach using variable length Markov models and taking as observations the contour of a human body in terms of control points has been developed to learn and recognise human activities, and has been evaluated in exercise activities that require no object interactions [20, 21].”
A survey on visual surveillance of object motion and behaviors (2004, background): “Active contour-based tracking algorithms track objects by representing their outlines as bounding contours and updating these contours dynamically in successive frames [6], [71], [72], [74].”
AI Choreographer: Music Conditioned 3D Dance Generation with AIST++ (2021, methodology): “Earlier works employ statistical models such as kernel-based probability distribution [64, 10, 25, 11] to synthesize motion, but abstract away motion details.”
Discriminative subvolume search for efficient action detection (2009, background): “Another type of approaches uses sequences of silhouettes or body contours to model actions [8] [14] and it requires background subtraction.”
Understanding Video Events: A Survey of Methods for Automatic Interpretation of Semantic Occurrences in Video (2009, methodology): “Variablelength Markov models (VLMMs) [80], [81] calculate the optimal level of temporal dependence using a divergence crite-”
Motion synthesis from annotations (2003, methodology): “The resulting Markov chain can be searched using dynamic programming to find a motion that connects two keyframes [Molina-Tanco and Hilton 2000] or used in a variable length Markov model to infer behaviors [Galata et al. 2001] or directly sampled from to create new motions [Bowden 2000].”