In this paper, we assess the capability of a uniqueunobtrusive footprint imaging sensor system, based on plasticoptical fiber technology, to allow efficient gait analysis fromtime domain sensor data by pattern recognition techniques. Trialgait classification experiments are executed as ten manners ofwalking, affecting the amplitude and frequency characteristics ofthe temporal signals. The data analysis involves the design of fivetemporal features, subsequently analyzed in 14 different machinelearning models, representing linear, non-linear, ensemble, anddeep learning models. The model performance is presentedas cross-validated accuracy scores for the best model-featurecombinations, along with the optimal hyper-parameters for eachof them. The best classification performance was observed for arandom forest model with the adjacent mean feature, yieldinga mean validation score of 90.84\%±2.46\%. We conclude thatthe floor sensor system is capable of detecting changes in gaitby means of pattern recognition techniques applied in the timedomain. This suggests that the footprint imaging sensor systemis suitable for gait analysis applications ranging from healthcareto security.