Login / Signup

A deep dive into the use of local positioning system in professional handball: Automatic detection of players' orientation, position and game phases to analyse specific physical demands.

Thomas LefèvreBrice GuignardClaude KarcherXavier RecheRoger FontJohn Komar
Published in: PloS one (2023)
The objective of this study is to automate and analyse the quantification of external load during an elite men's handball match. This study was carried out using data from a local positioning system and inertial measurement units. The literature review leads us to assume that physical demands are different depending on position, player specialty and phases of the game. In order to do this analysis, raw data was used from professional competitors of a Spanish club during National and European competition matches. First, a game phase algorithm was designed to automate phase recognition. Then, a descriptive evaluation of the means and standard deviation was performed with the following variables: total distance, total time, total Accel'Rate, the percentages of distance and time per speed and displacement direction. A Kruskal Wallis test was applied to normalized distance and normalized Accel'Rate. Defensive play showed the highest values on covered distance (930.6 ± 395.0 m). However, normalized distance showed significant differences (p<0.05) across all phases with defensive play (558.8 ± 53.9 m/10min) lower than offensive play (870.3 ± 145.7 m/10min), offensive transition (1671.3 ± 242.0 m/10min) or defensive transition (1604.5 ± 242.0 m/10min). Regarding position, wing players covered the most distance (2925.8 ± 998.8 m) at the second highest intensity (911.4 ± 63.3 m/10min) after offensive back players (1105.0 ± 84.9 m/10min). Significant difference in normalized requirements were found between each playing position: goalkeepers, wings, versatile backs, versatile line players, offensive backs and defensive backs (p<0.05), so a separation between offensive or defensive specialists is plausible and necessary. In conclusion, as physical demands differ for each game phase, activity profile among players is modulated by their playing position and their specialty (offense, defense or none). This study may help to create individual training programs according to precise on-court demands.
Keyphrases
  • middle aged
  • physical activity
  • mental health
  • machine learning
  • deep learning
  • body composition
  • big data
  • electronic health record
  • quality improvement
  • artificial intelligence
  • mass spectrometry
  • sensitive detection