In this study a fuzzy logic classification system was used first to discriminate healthy subjects from patients rather than classifying those using Brunnstrom stages. Decision making was performed in two stages: feature extraction of gait signals and the fuzzy logic classification system which is used Tsukamato-type inference method. According to our signal feature extraction studies, we focused on temporal events and symetrical features of gait signal. Developed system has six inputs while four of them for temporal features evaluation rule block and two of them symmetrical features evaluation rule block. Our simulation test results showed that proposed system classify correctly 100% of subjects as patient and healthy elderly. The correlation coefficient was found 0.85 for classification to subjects to correct Brunnstrom stages. The results show that classifying patients becomes increasingly difficult linearly according to hemiplegia’s severity.