As the number of bicyclists grows steadily, bicycle related crashes are getting more concerns. In order to enhance bicycle safety, various factors affecting bicycle crash risk have been investigated. This study attempts to extract the kinematic and the spatial features from a large number of the bicycle trajectories and to investigate their significance as the crash risk factors. The trajectories were collected on the segregated, i.e., free from motor vehicle traffics, two-way riverside bicycle track by using GPS units. As the kinematic features, the speed and the heading change were estimated and as the spatial ones, several distance measures were evaluated on two summarized trajectories of the GPS data in both directions. The summary trajectories were constructed using the principal curve method. The extracted features were applied to the regression analysis in order to confirm their correlation to the bicycle crash risk. Our results show that the kinematic features were rather irrelevant while the spatial ones had meaningful correlation with the crash risk and thus confirm their usefulness in crash risk analysis.