Einleitung

Improving one’s skills in a sport such as snowboarding, which requires complex motion sequences and perfect balance, is often only possible through expensive lessons (in Switzerland currently around 80 EUR per hour) given by an experienced teacher (Primus, 2007). Yet even this approach has its limitations, as instructors might not notice all mistakes. We therefore envision and have been developing SnowPro, a wearable sports trainer capable of supporting snowboarders in improving their skills (Holleczek et al., 2009). One feature of SnowPro is the analysis of the dynamics of the weight distribution inside the boots with the help of pressure sensors integrated into a pair of insoles. This is essential for the recognition of wrong weight shifting techniques, which usually lead to painful crashes. Usually, the basis of a wearable sports trainer is a sensor system recognizing the activities performed by its user. In a second step, the segmented activities are evaluated regarding specific quality measures. Finally, a feedback component informs the user (in real-time or after the workout) about her performance and supports her during the learning process. In the snowboarding domain, the basic user activities to be recognized by the wearable sports trainer are turns of the riders. Traditional activity recognition approaches make use of pressure insoles with force-sensitive resistors (FSR5), which, however, are particularly uncomfortable to wear (Holleczek et al. 2009). In this paper, we address the following research questions: 1. Can we integrate FSRs into conventional snowboard bindings to make pressure measurements more convenient? 2. Are these modified snowboard bindings capable of withstanding the hostile conditions on outdoor ski slopes? 3. Can the obtained sensor data be used to detect turns, the basic snowboarding activity, with the help of a simple machine learning algorithm? Einleitung