| dc.description.abstract | Background: Data models have been developed to recognize common physical activity types (e.g., walking, standing, sitting) based on accelerometer. However, few data models have been developed to investigate whether complex physical activity (CPA) can be identified based on the composition of the activity types. To develop data models that manage to identify CPA in adolescents, it is necessary to create a valid and firm base of activity definitions. However, there are no common definitions or structures on human activities that allow us to formulate clear and precise activity definitions.
Aim: The purpose of this pilot study is twofold. First, to develop a set of activity definitions that can be used to classify vigorous physical activity. Second, to explore time distribution and characteristics of activity types and movements during CPA. Activity and movements during ballgames (handball and football) is used as paradigm example.
Methods: Six participants carried out a training session of football and six participants carried out two training sessions of handball. The participants were filmed with four GoPro cameras that were placed in each corner of the court, to record the ballgames. A set of 32 activity definitions were developed to guide the annotation of the video recordings. The videos were later annotated frame-by-frame and used to calculate the time distribution of activity types (i.e., total time per activity and average duration per event for each activity). Inter-rater reliability (IRR) of the video annotation was calculated in ANVIL.
Results: The IRR from the video annotation were 0.87 (Cohen's kappa coefficient) between the two coders. A total of 14 h, 14 min and 59 s was video annotated for 11 participants, one participant was excluded. About 27% of the annotated video time was classified as “undefined”. The most common activities were “walking forwards”, “shuffling”, “standing”, “running forwards” and “sitting”. In sum, these activities occupied 63% of the training sessions. The average duration was <2% for separate instances of different activities, such as “kick” and “skipping sideways”.
Conclusion: Static postures and activities that typically are performed in cyclical movement patterns (e.g., running and walking), occupy more time than discrete movements (e.g., kicking, throwing) during ball games activity. When excluding “undefined”, activities such as “walking forwards”, “shuffling”, “standing” and “running forwards” got the highest average duration compared to the rest of the activity types during both of the ball games. | nb_NO |
