Classification of the Stride Length based on IMU Sensor using the Decision Tree

Abstract

One of the factors to be measured during stroke patients’ walking training is their step length. Stride length can be assessed based on the distance from heel to heel and can then be used to identify whether patient steps are short, medium, or long. This investigation offers to classify the length of the inventory patient by employing an inertial measurement unit (IMU) sensor. The patient leg is fitted with an IMU sensor to assess the leg position. In this study. Every 0.01 second, the IMU sensor data including accelerometer and gyroscope are read and processed using many stages including signal pretreatment, extraction of features and the classification models. To retrieve the signal feature, the signal read by the IMU sensor is then preprocessed and extracted. The characteristics extracted are the amplitude of the negative peak, the amplitude of the positive peak and time from the negational to the positive peak. The feature is then used as the input for the decision tree to categorize the stride length as short, medium, or long stride. The categorization error is 6.64% and can be classified as a low error. The proposed model can therefore be used to classify the length of the steps while walking.