Parkinson's disease (PD) is a neurodegenerative disorder. With progression of PD, movement disorder such as gait disturbance and balance impairment is frequently observed. Hoehn and Yahr scale (HY) is an indicator to evaluate the severity of motor signs of PD. Recently, objective measurement comes to be widely spread. Previous studies pointed out that human walking comes from complex interaction, and it comes to be seen as nonlinear dynamics. Amplitude of PD patients' stride time variability was reported to be larger than that of healthy people. Coefficient of variation (CV) is commonly used to see amplitude of variability. The fractal property of PD patients' stride time is lower than that of healthy people. The fractal property was measured by scaling exponent alpha calculated by detrended fluctuation analysis (DFA). However, the relationship between the stride time variability and the severity of motor signs of PD remains to be clarified. In this study, we tried to investigate the relationship between these indicators. Clarifying the relationship between practical severity index and objective data provides us information to make a diagnosis of PD. Forty-five PD patients walked 200 meter corridor at their preferred pace. As control group, 35 healthy people, which include young and elderly people, are participated. In order to separate between the PD patients and healthy people or to classify HY scale, linear discriminant analysis on both CV and DFA was applied. When we separated into all PD patients group and all healthy people group, the accuracy was 0.76. We tried to separate 2 groups. One is group of PD patients with HY2.5 or higher and the other is the group of healthy elderly people or mild PD patients. The specificity was 0.86. When we tried to separate between PD patients with HY3 and PD patients with HY2.5, the sensitivity was 0.93. We conclude that gait variability is one of the indicators of motor severity seen in PD.