Using Dynamics to Recognize Human Motion

Venture, Gentiane; Yabuki, Takumi; Kinase, Yuta; Berthoz, Alain; Abe, Naoko

doi:10.1007/978-3-319-25739-6_17

Cited by 3 publications

(2 citation statements)

References 29 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The combination of dimensionality reduction and machine learning techniques is not new in the field of motor behavior. It has been adopted for the analysis of simple reaching tasks (Ansuini et al 2016), for the analysis of whole body kinematics both in the field of human-computer interaction (Venture et al 2016;Zhang and Venture 2012) and in the fields of biomechanics and sport science (Huys et al 2008). The main novelty of our approach consists in the use of the spatiotemporal PCA decomposition for representing the whole body kinematics of complex motor tasks, which are shown to support remarkably compact representations of the complex kinematics characterizing noncyclic actions.…”

Section: Discussionmentioning

confidence: 99%

A whole body characterization of individual strategies, gender differences, and common styles in overarm throwing

Maselli

Dhawan

Russo

et al. 2019

Journal of Neurophysiology

View full text Add to dashboard Cite

Overarm throwing is a fundamental human skill. Since paleolithic hunter-gatherer societies, the ability of throwing played a key role in brain and body co-evolution. For decades, throwing skill acquisition has been the subject of developmental and gender studies. However, due to its complex multijoint nature, whole body throwing has found little space in quantitative studies of motor behavior. In this study we examined how overarm throwing varies within and between individuals in a sample of untrained adults. To quantitatively compare whole body kinematics across throwing actions, we introduced a new combination of spatiotemporal principal component, linear discrimination, and clustering analyses. We found that the identity and gender of a thrower can be robustly inferred by the kinematics of a single throw, reflecting the characteristic features in individual throwing strategies and providing a quantitative ground for the well-known differences between males and females in throwing behavior. We also identified four main classes of throwing strategies, stable within individuals and resembling the main stages of throwing proficiency acquisition during motor development. These results support earlier proposals linking interindividual and gender differences in throwing, with skill acquisition interrupted at different stages of the typical developmental trajectory of throwing motor behavior. NEW & NOTEWORTHY Unconstrained throwing, because of its complexity, received little attention in quantitative motor control studies. By introducing a new approach to analyze whole body kinematics, we quantitatively characterized gender effects, interindividual differences, and common patterns in nontrained throwers. The four throwing styles identified across individuals resemble different stages in the acquisition of throwing skills during development. These results advance our understanding of complex motor skills, bridging the gap between motor control, motor development, and sport science.

show abstract

Section: Discussionmentioning

confidence: 99%

A whole body characterization of individual strategies, gender differences, and common styles in overarm throwing

Maselli

Dhawan

Russo

et al. 2019

Journal of Neurophysiology

View full text Add to dashboard Cite

show abstract

“…Thus, the designation of body movements and their emotional association should primarily be driven by the dancers and choreographers. With such knowledge, it is possible to design a system reacting to perceived emotions by categorizing the body motion (Venture et al 2016).…”

Section: Expressive Motionmentioning

confidence: 99%

Engaging with Robotic Swarms

St-Onge

Côté-Allard

Glette

et al. 2019

J. Hum.-Robot Interact.

View full text Add to dashboard Cite

In recent years, researchers have explored human body posture and motion to control robots in more natural ways. These interfaces require the ability to track the body movements of the user in three dimensions. Deploying motion capture systems for tracking tends to be costly and intrusive and requires a clear line of sight, making them ill adapted for applications that need fast deployment. In this article, we use consumer-grade armbands, capturing orientation information and muscle activity, to interact with a robotic system through a state machine controlled by a body motion classifier. To compensate for the low quality of the information of these sensors, and to allow a wider range of dynamic control, our approach relies on machine learning. We train our classifier directly on the user to recognize (within minutes) which physiological state his or her body motion expresses. We demonstrate that on top of guaranteeing faster field deployment, our algorithm performs better than all comparable algorithms, and we detail its configuration and the most significant features extracted. As the use of large groups of robots is growing, we postulate that their interaction with humans can be eased by our approach. We identified the key factors to stimulate engagement using our system on 27 participants, each creating his or her own set of expressive motions to control a swarm of desk robots. The resulting unique dataset is available online together with the classifier and the robot control scripts.

show abstract