Constantin Zwick

Your personal movie producer: generating highlight videos in soccer using wearables

Manually browsing through high amount of sports videos, selecting interesting highlight scenes, and applying video effects are time-consuming and one major burden these days. Automatic approaches are preferred, but currently require high-quality TV broadcast material which is usually not available in recreational sports. Thus, the purpose of this paper was to develop a personal low-cost movie producer for highlight videos using wearables. The feasibility of the proposed approach was shown for soccer scenarios. The automatic highlight video generation included three contributions: (i) sensor-based full-instep kick detection and extraction of corresponding video segments, (ii) sensor-based ball speed estimation for provision of highlight-related metadata shown in the final video, and (iii) sensor-driven video effect generation. The proposed system was evaluated on eleven subjects which were equipped with inertial sensors in the cavity of soccer shoes. A mean sensitivity of 95.6 % and a mean absolute error of 7.7 km/h were achieved for the full-instep kick classification and the ball speed estimation, respectively. This personal movie producer based on wearables is a novel idea to provide recreational athletes with attractive automatically generated highlight videos in sports.