As the field of medicine continues to develop the life expectancy of the population will also continue to grow. This increase in life expectancy will lead to a greater occurrence of several possible pathologies at an advanced age. Some pathologies affect the nervous system which in turn can lead to a degradation of the patient’s gait and a decrease in quality of life.

Similar to the field of medicine, the development of gait monitoring systems has significantly increased the amount of data that can be extracted from a patient’s gait. With this increase in acquired data the time required to perform this analysis will also increase significantly as well as the probability of misdiagnosis. Machine learning algorithms can become a powerful tool in the diagnosis of several pathologies due to their capacity to process great amounts of data and uncover patterns in a patient’s gait that are not obvious to humans. These methods can significantly decrease the time required to analyze all the patient’s data leading to an automatic, time-effective and more accurate analysis.

There are currently several published studies related to automatic gait pattern recognition that use machine learning methods in which a classifier is developed to distinguish between the gait of a healthy subject and a pathological one. The majority of these studies fail to explore a quantitative relationship between the pathological symptoms in a patient’s gait and the changes observed in the gait data acquired. Furthermore, most of these studies only implement binary classification in which each training set can only be associated to one of two possible classes. In the case of multi-class classification there are more than two possible classes. The few studies that do make use of multi-class classification experience a drop in performance when the classifier is set to distinguish between different stages of the same pathology.