Many lower extremity orthoses for preventing stumbling have been developed until now. However, by the conventional method, the orthosis cannot predict stumbling while assisting motion. Therefore, it need always supporting. But, if it always supports user, there is a possibility of blocking an ability of user's independent walk. So, the purpose of our research is to develop the lower extremity orthosis which can support only in the case of stumbling. In this paper, the objective is to develop the algorithm for predicting stumbling and decision of the requirement specification about orthosis in case of using the algorithm. We thought that we can judge the user will stumble or not from comparing the data of currently performed and usually performed because the walk of man was pattern movement. Therefore, we decided to predict stumbling by using significance test about the motion data of the leg at upping step. In our experiment, the motion data at upping step of paraplegia was measured using three dimensional motion analysis system. When adapted for measurement data in the proposal algorithm, it was shown that predicting stumbling is possible. Moreover, because applicability for healthy person is also checked by exploratory experiment, the proposal algorithm is also expected to be able to apply for the user of the wide range condition. However, it is required to increase a subject and to verify the further applicability.