How data science can catalyze scientific support to performance? 

Swimming Monitoring & Modelling of Performance : how data science can catalyze scientific support to performance ?

In the age of the digital revolution, swimming is still under-technologized. However, the current densification of the top-level sport is forcing ever greater optimization of the training process, and detailed analysis of competitions. The contribution of data-driven approaches, through the routine use of inertial measurement unit, provides for sport science an interesting perspective on in-situ analysis during swimming.

However, exploiting these data for scientific support to performance poses a number of obstacles, especially methodological ones, in order to extract relevant insights from a multivariate time series relating to 3D accelerations and angular velocities of swimmers. 

This presentation will give an overview of such works including, on the one hand, methodological contributions based on deep learning for biomechanical monitoring (i.e., human activity recognition) and functional clustering for technical skills profiling of swimmers. On the other hand, practical applications with the French Swimming Federation and Olympic swimmers will be briefly presented.