Project Detail

Disentangling duelling Judoka: Automated performance analysis in combat sports

Aangemaakt door

David Mann

Website

www.fgb.vu.nl

Contact

d.mann@vu.nl

Start Datum

2016-03-01

Eind datum

2016-09-30

Wedstrijdanalyse in vechtsport wordt momenteel veelal handmatig uitgevoerd. Dit is een moeizaam en tijdrovend proces, waardoor de toepassing beperkt blijft tot de topsport. Het ultieme doel van dit project is om software te ontwikkelen die de bewegingen van vechtsporters tijdens een gevecht automatisch in kaart kan brengen. Coaches en sporters op alle niveaus kunnen daarmee inzicht krijgen in het wedstrijdverloop en beslissende acties die leiden tot winst of verlies. Wij maken hierbij gebruik van de Kinect sensor ontwikkeld door Microsoft. Deze sensor kan bewegingen van personen op basis van infraroodlicht driedimensionaal in kaart te brengen. De Kinect kan meerdere personen van elkaar onderscheiden zolang deze zich op enige afstand van elkaar bevinden, maar wanneer zij te dicht bij elkaar komen dan gaat het mis. De eerste stap is het ontwikkelen van een prototype waarmee twee judoka’s die in gevecht zijn uit elkaar gehouden kunnen worden en hun individuele bewegingen gevolgd.

Value Proposition

This idea will lead to the development of automated performance analysis software for use in combat sports (e.g., judo, karate, boxing, taekwondo) using Microsoft Kinect sensors. These sensors are not presently able to distinguish two opponents who are in contact with each other (as is the case in some combat sport), and so in this project we will develop algorithms to (1) distinguish two opponents on the basis of footage from Kinect sensors, and (2) categorise particular actions performed by the two opponents.

Revenue Model

The multi-million euro market for performance analysis software in sport (e.g., Catapult, Dartfish) is large and growing rapidly. Combat sports are presently required to use generic software packages and adapt them to their sport. Given the number of combat sports included in, for example, the Olympic and Paralympic Games (and therefore the number of medals on offer), it is expected that there would be very strong interest in this software from all countries looking for a competitive edge in these competitions.

Tooling, Methods, Data

The project requires Microsoft Kinect sensors to record three-dimensional depth images of judo bouts. Computers with high-end graphic cards will be used to process the images by identifying and distinguishing the opponents, and then will use algorithms designed to identify basic actions.

Expertise Knowledge

1. Programming expertise to write computer algorithms capable or distinguishing two separate opponents when in contact with each other. 2. Sport-specific technical expertise to identify particular actions to be identified and coded.

Confidentiality

The algorithms written as part of the project will remain confidential.

Big Data Factor

The long-term aim is to develop a system that can be used for a range of different combat sports to not only provide specific information about opponents, but also to combine information centrally to facilitate a ‘big data’ approach to performance analysis in these sports. Computer-driven pattern analysis and recognition could be used to reveal the previously un-thought-of sequences of movements and behaviour most likely to lead to success later in a bout.

Organizations and People Involved

Vrije Universiteit Amsterdam (Bert Coolen, David Mann, Rianne Ravensbergen, Kai Krabben, John van der Kamp) - development and pilot testing of algorithms. Top Judo Amsterdam (John van der Kamp) - provide facilities and athletes for recording judo footage. Judo Bond Nederland (Richard Smit) - access to existing performance analysis data on judo to identify particular judo actions to prioritise.

Key Success Factors

By the end of this SportInnovator idee our aim is to: 1. Modify existing software to create a full depth profile of two athletes (rather than one person) 2. Develop an algorithm that distinguishes the two athletes from each other using the depth profile from the Kinect sensor 3. Develop an algorithm that identifies the prioritised actions: (i) when the two judoka are in contact with each other, and (ii) when the judoka are standing or on the ground.

Target Customer Audience

End users will be sport organisations that cater for combat sports (e.g., National Federations, International Federations, academies, training centres).

Solution Summary

The aim of this idee is to develop prototype performance analysis software capable of automatically detecting and coding basic actions performed in a combat sport. The long-term vision is to develop commercially available software capable of performing automated performance analysis across a range of different combat sports.

Kalender

Gerelateerd Nieuws

Gerelateerde Projecten