Bran-machine interface (BMI) can be used for controlling of external devices such as the exoskeleton, robot arm, etc. For efficient communication between a user and machine, fast and accurate detection of user intention is important elements in the BMI application. For this reason, readiness potential (RP) is a useful feature that is possible to detect movement intention before the movement onset. To our knowledge, however, the analysis of single-Trial RP component has not been sufficiently investigated in the real-world application (e.g. powered exoskeleton or robot arm). In our study, we first validate a single-Trial RP performance in the lower limb exoskeleton environment where the user allows for voluntary walking. The experiments are executed in the two different walking conditions which are normal and exoskeleton walking. The Laplacian and common average reference (CAR) filters are applied to reduce spatial noise and regularized linear discriminant analysis (RLDA) is used as a classifier. Our results show the averaged classification accuracy of80.7% for 5 subjects. This study demonstrates a feasibility of RP-based BMI system for controlling of a lower limb exoskeleton.