Front crawl is an alternating swimming stroke technique in which different phases of arm movement induce changes in acceleration of limbs and body. This study proposes a new approach to use inertial body worn sensors to estimate main temporal phases of front crawl. Distinctive features in kinematic signals are used to detect the temporal phases. These temporal phases are key information sources of qualitative and quantitative evaluation of swimming coordination, which have been assessed previously by video analysis. The present method has been evaluated upon a wide range of coordination and showed a difference of 4.9% with video based system. The results are in line with video analysis inter-operator variability yet offering an easy-to-use system for trainers.