Gait asymmetry as a consequence of hemiparesis is known as a serious long-term disability, where typical compensatory gait movements are used by stroke subjects in order to cope with their daily activities. This study presents a gait symmetry training approach consisting of the adaptive robot assistance with the combination of the visual feedback integrated in Balance Assessment Robot for treadmill walking (BAR-TM). The support algorithm is based on gait temporal parameters that change the level of assistive support, which was triggered at different gait subphase events. The presented approach was evaluated on a post-stroke subject in a large number of training sessions. The results have shown significant improvement of gait symmetry indicating that the proposed rehabilitation method has significant potential that should be explored in further studies.