Fatigue Level Prediction of Lower Extremity Knee Flexor and Extensor Muscles using Neural Network

The purpose of this paper is to build a model to predict the fatigue level of lower extremity knee flexor and extensor muscles based on sensor data. We used maximum voluntary contraction (MVC) to assess the fatigue level of muscles. We collected walking data from six healthy adults at four MVC levels. These data come from sixteen pressure sensors on the insoles and two inertial measurement units (IMUs) on the two feet. Fifty-one gait features were extracted from these data. Through using the ReliefF Algorithm, the fourteen features most relevant to the MVC level were extracted. The accuracy of neural network classifier is 72.7%. This indicates that the selection of features and the use of neural networks is reasonable. This study used fewer sensors and predicted more subdivided fatigue levels than previous studies. Findings from this study can help better understand how fatigue of knee flexor and extensor muscles affect gait characteristics, and further aid in developing an early warning system to prevent knee injuries.