The main aim of this paper is to explore application of fuzzy rules for automated recognition of gait changes due to falling behaviour. Minimum foot clearance (MFC) during continuous walking on a treadmill was recorded on 10 healthy elderly and 10 elderly with reported balance problem and tripping falls. MFC histogram characteristic features were used as inputs to the set of fuzzy rules; the features were extracted based on estimating the clusters in the data. Each of the clusters found corresponded to a new fuzzy rule, which were then applied to associate the input space to an output region. Gradient descent method was used to optimise the rule parameters. Both cross-validation and Jack-knife (leave-one-out) techniques were utilized for training the models and subsequently, testing the performance of the optimized fuzzy model. Receiver operating characteristics (ROC) plots, as well as accuracy rates were used to evaluate the performance of the developed model. Test results indicated up to a maximum of 95% accuracy in discriminating the healthy and balance-impaired gait patterns. These results suggest good potentials for fuzzy logic to use as gait diagnostics.