Gait analysis plays important role for detecting the medical condition of a person and applying to make rehabilitation plans for disabled people. Recently, a gait analysis system 　which recognizes each footprint from foot trace images recorded on a paper of 10m length, and calculates various parameters such as each step length, step width, etc. has 　been developed. In this study, we measure and evaluate the walking behavior of hemiplegic patient by using the gait analysis system. The change of gait parameters according to the rehabilitation progress and the relation to the rehabilitation program are analyzed statistically. As a result, it is shown that the frequency and length of shuffle and the walking straightness computed by the mid-point of step width are useful indices to determine the recovery degree of patient.