Incremental Learning, Clustering and Hierarchy Formation of Whole Body Motion Patterns using Adaptive Hidden Markov Chains

By Dana Kulic, Wataru Takano, Yoshihiko Nakamura. 2008

233 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering, Social Psychology

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

1. Luc Van Gool
2. Stefan Wermter
3. German Ignacio Parisi
4. Pieter Abbeel
5. Jan Peters
6. Aude Gemma Billard
7. Luiza Mici
8. Sylvain Calinon

In Control and Systems Engineering

1. Jan Peters
2. Pieter Abbeel
3. Aude Gemma Billard
4. Sergey Levine
5. Gerhard Neumann
6. J. Bagnell
7. Sylvain Calinon
8. Seyed Mohammad Khansari-Zadeh

In Social Psychology

1. Aude Gemma Billard
2. Sylvain Calinon
3. Darwin G. Caldwell
4. Seyed Mohammad Khansari-Zadeh
5. Pieter Abbeel
6. German Ignacio Parisi
7. Michael Beetz
8. Stefan Wermter

Citations in the corpus

• An Incremental Self-Organizing Architecture for Sensorimotor Learning and Prediction (2017, methodology): “[29] used HMMs for segmenting and representing motion patterns together with a clustering algorithm that learns in an incremental fashion based on intramodel distances.”
• An Algorithmic Perspective on Imitation Learning (2018, methodology): “Recognition with HMMs works well when the number of states is relatively low [Kulić et al., 2008].”
• BM: An iterative algorithm to learn stable non-linear dynamical systems with Gaussian mixture models (2010, background): “Modeling point-topoint motions thus provides basic components, so-called motion primitives, for robot control [1], [2].”
• Learning and Reproduction of Gestures by Imitation (2010, methodology): “As an alternative, other approaches have considered modeling the intrinsic dynamics of motion [13]–[16].”
• Learning Stable Nonlinear Dynamical Systems With Gaussian Mixture Models (2011, methodology): “Modeling point-to-point motions provides basic components for robot control, whereby more complex tasks can be decomposed into sets of point-to-point motions [1], [2].”
• Learning Physical Collaborative Robot Behaviors From Human Demonstrations (2016, background): “HMMs that were incrementally updated during reproduction [26].”
• Incremental learning of full body motion primitives and their sequencing through human motion observation (2012, methodology): “Segmented motion primitive candidates are incrementally clustered to abstract generative models of the motion primitives (Kulić et al. 2008a).”
• The KIT Motion-Language Dataset (2016, unknown): “should be annotated next if not all motions have been annotated yet or if new motion data is added. A possible approach could be to use unsupervised clustering using Hidden Markov Models (HMMs) [32], [33]. For an unannotated motion, the distance to existing clusters could be used to decide how to pr”
• On Human Action (2011, background): “Many authors define the set of primitives by hand while others employ statistical learning schemes [27, 28].”
• The TUM Kitchen Data Set of everyday manipulation activities for motion tracking and action recognition (2009, methodology): “Data is provided in the BVH file format6, which contains the 6 DOF pose and the joint angle values as defined in Fig.”
• Task-Specific Generalization of Discrete and Periodic Dynamic Movement Primitives (2010, unknown): “[24], [25] extended these works by show-”