In this paper, a novel gait assessment system based on measures of gait variability reflected through the variability of shapes of gait cycles trajectories is proposed. The presented gait assessment system is based on SVM (support vector machine) classifier and on gait variability-based features calculated from the hip and knee joint angle trajectories recorded using wearable IMUs during walking trials. A system classifier was trained to distinguish healthy gait patterns from the pathological ones. The features were extracted by calculating the distances between the joint trajectories of the individual gait cycles using 4 different distance functions. As result, the system is able to provide a Gait Variability Index (GVI), which is a numeric value that can be used as an indicator of a degree to which a pathological gait pattern is close to a healthy gait pattern. The system and GVI were tested in three experiments, involving subjects suffering from gait disorders caused by different neurological diseases. The results demonstrated that the proposed gait assessment system would be suitable for supporting clinicians in the evaluation of gait performances during the gait rehabilitation procedures.