Velocity-Based Multiple Change-Point Inference for Unsupervised Segmentation of Human Movement Behavior

Senger, Lisa; Schröer, Martin; Metzen, Jan Hendrik; Kirchner, Elsa Andrea

doi:10.1109/icpr.2014.781

Cited by 15 publications

(11 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

Section: Learning Platformmentioning

confidence: 98%

“…These borders can be determined using an online Viterbi-algorithm in a fully unsupervised manner. By including the velocity into the inference in the vMCI algorithm, segments with a bell-shaped velocity can be detected automatically without need for parameter tuning, as shown in several experiments (Senger et al, 2014;Gutzeit and Kirchner, 2016). Because in general there is no ground truth segmentation available for human movements, the vMCI algorithm has been compared to other segmentation algorithms on synthetic data generated from sequenced dynamical movement primitives [DMP; (Ijspeert et al, 2013)] as well as on real human manipulation movements (Senger et al, 2014).…”

Section: Behavior Segmentationmentioning

confidence: 99%

“…Manipulation movements usually have bell-shaped velocity profiles ( Morasso, 1981 ). Based on this knowledge, we developed the velocity-based Multiple Change-point Inference (vMCI) algorithm for the segmentation of manipulation behavior ( Senger et al, 2014 ). The vMCI algorithm is an extension of the Multiple Change-point Inference presented by Fearnhead and Liu (2007) , in which Bayesian inference is used to determine segment borders in a time series.…”

Section: Learning Platformmentioning

confidence: 99%

“…The vMCI algorithm is an extension of the Multiple Change-point Inference presented by Fearnhead and Liu (2007) , in which Bayesian inference is used to determine segment borders in a time series. We extended this algorithm to detect building blocks in human manipulation movement ( Senger et al, 2014 ). Each segment is represented with a linear regression model.…”

Section: Learning Platformmentioning

confidence: 99%

See 3 more Smart Citations

The BesMan Learning Platform for Automated Robot Skill Learning

et al. 2018

Self Cite

View full text Add to dashboard Cite

We describe the BesMan learning platform which allows learning robotic manipulation behavior. It is a stand-alone solution which can be combined with different robotic systems and applications. Behavior that is adaptive to task changes and different target platforms can be learned to solve unforeseen challenges and tasks, which can occur during deployment of a robot. The learning platform is composed of components that deal with preprocessing of human demonstrations, segmenting the demonstrated behavior into basic building blocks, imitation, refinement by means of reinforcement learning, and generalization to related tasks. The core components are evaluated in an empirical study with 10 participants with respect to automation level and time requirements. We show that most of the required steps for transferring skills from humans to robots can be automated and all steps can be performed in reasonable time allowing to apply the learning platform on demand.

show abstract

Section: Learning Platformmentioning

confidence: 98%

Section: Behavior Segmentationmentioning

confidence: 99%

Section: Learning Platformmentioning

confidence: 99%

Section: Learning Platformmentioning

confidence: 99%

See 2 more Smart Citations

The BesMan Learning Platform for Automated Robot Skill Learning

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…For this decomposition of behavior, unsupervised machine learning methods for behavior segmentation can be applied [74]. For instance, human demonstrations of complex behavior can be broken down into simpler behavioral building blocks which are easier to learn by imitation and more reusable.…”

Section: Improved Interaction Based On Learning From Humansmentioning

confidence: 99%

Formal Modeling and Verification of Cyber-Physical Systems

Drechsler¹,

Kühne²

2015

View full text Add to dashboard Cite

rial is concerned, speci cally the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on micro lms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a speci c statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.Printed on acid-free paper Springer Vieweg is a brand of Springer Fachmedien Wiesbaden Springer Fachmedien Wiesbaden is part of Springer Science+Business Media (www.springer.com) PrefaceToday, embedded systems are ubiquitous in our everyday life, in cell phones and washing machines, but also in cars and even medical equipment. With this escape from their former habitat within desktop computers, these systems are increasingly interacting with their environment: Sensors measure physical phenomena such as temperature, acceleration, or magnetic fields, while actors manipulate the outside world, like in robots or electronically controlled combustion engines. The combination of an electronic system with a physical process is called a cyber-physical system (CPS).With this paradigm, many new challenges need to be faced during modeling, design, implementation, verification, and test. For the design of hardware and software of CPS, new approaches need to be developed, taking into account non-functional requirements like energy efficiency and reliability even in harsh environments. Real-time aspects often play an important role. Furthermore, if a system is interacting with its physical environment, it becomes difficult to prove the functional correctness of the system. The combination of discrete and continuous behavior and the treatment of noisy sensor data are challenging problems.Considering all the above aspects, CPS is an interdisciplicary topic, touching research areas such as computer science, robotics, electrical and mechanical enginieering, physics and more. In the First International Summer School on Methods and Tools for the Design of Digital Systems, we want to bring together experts from different fields of research and application. The goal of this summer school is to introduce PhD students to this exciting topic, and to offer them deeper insights into formal modeling and verification techniques. Regarding the application areas, the courses focus on robotics and space systems, as well as the railway domain and microfluidic biochips.This s...

show abstract