Summary form only given. Human gait analysis can be performed by using a treadmill and two aligned web cameras, positioned one on each side of the treadmill. In this system, passive marks are positioned on person’s joints and various angles of the gait are recorded by the cameras at different speeds of the treadmill. The treadmill’s speed is appropriated for each person clinical case. This system is a substantial evolution from  at a much lower cost than  and . This research project aims to create software capable of generating joint trajectory references of healthy people gaits, considering height, weight, age and test speed. These trajectories will be used as reference to compare with the data of a person with an abnormal gait. From this comparison a classification of the severity of the pathology will be obtained. The developed software uses an artificial neural network, based on 97 samples from 20 walking people with healthy gaits, collected on treadmill’s tests. 70% of the samples were used for training, 5% for validation and 25% for testing.
The two best neural networks for the knee joints are constituted by 10 or 12 neurons in the hidden layer, showing regression values higher than 97%. They have four inputs (height, weight, age and test speed) and the output is the reference knee joint trajectory. In this project it is also used the extreme learning machine, as an alternative computational intelligence approach of the neural network. With this software physiotherapists can make gait pattern comparisons taking into account the specific characteristics of each person, instead of comparisons with the standard gait patterns of the literature that does not differentiate for different characteristics. The system was tested analyzing the gait of 7 persons who were subjected to ligamentoplasty (surgical reconstruction) about two years ago, after suffering a rupture of the anterior cruciate ligament of the knee. Collected data were compared with the traj- ctory references generated by the software for each person taking into account their physical characteristics. The results show that this software makes it possible to analyze and quantify the severity of gait pathologies, which is a significant improvement to the present subjective analysis practice.