Inertial Sensors to Detect Multiple Gait Disorders

Pathological gait analysis is of vital importance in clinical applications. In this paper, a study on the development of a gait analysis system, based on three inertial sensors mounted on one of the shoes, is introduced. The proposed system can detect multiple disorders in real-time by performing gait analysis on the data collected from these inertial accelerometers. A novel feature selection approach is proposed which effectively filters out featureless information and extracts salient features in each gait type which forms a compact feature dataset for training. It is shown that the training of machine learning models using the transformed feature dataset is faster and more effective. The proposed system achieved an average of 100% identification accuracy of gait types from the transformed feature dataset.