We have been developing a robotic split-belt treadmill with different belt velocities and visual biofeedback of stance phase for the stroke patients to improve the symmetricity of their gait balance. In this treadmill system, a visual bio-feedback system of stance phase time of affected and unaffected legs is used. In this paper, a signal processing to eliminate spike and high frequency noises and oscillation of judge result of gait phase is discussed to indicate the stance phase time in real-time accurately. In addition, the effect of visual biofeedback of gait balance on their gait was analyzed in the determined biofeedback condition. Firstly, we conducted the gait experiment to determine the parameters in the signal processing such as cut-off frequency of low pass filter. In this experiment, the number of false judgment was calculated in normal gait, high cadence gait and simulated stroke gait. As a result, two parameters, cut-off frequency and the number of data to judge the gait phase, are set as 3.0 [Hz] and 10 respectively. In this condition, the gait phase of affected and unaffected legs are accurately displayed in front of stroke patient and physical therapist. Secondly, simulated stroke patients evaluated the effect of biofeedback of stance phase on the stance phase balance. As a result, the stance phase balance between affected and unaffected legs' stance phase was symmetrized by using visual biofeedback system.
