The Naïve Bayes Algorithm for the Stride Length Classification

Abstract

Stride length is one of the elements that will be examined for stroke patients' walking training. The stride length reveals the distance in stepping between the right and left feet or vice versa which can be used to track the development of a patient's training. This study proposes to use an inertial measurement unit (IMU) sensor to classify the step imbalance of a stroke patient. To monitor the position of the patient's leg, an IMU sensor is connected in the user thigh. The data from the IMU sensor is collected at a sample frequency of 100 Hz. The signal from the IMU sensor is then refined from the noise and exported to acquire the signal feature. In the investigation, the signal characteristic containing the positive and negative peaks of the accelerometer and gyroscope is employed. These four features are utilized as inputs for the Naïve Bayes model. The accuracy of the Nave Bayes model is 90.35 percent. The accuracy, recall, and F-measure of the Nave Bayes model are 0.906, 0.904, and 0.902, respectively. As a result, the Naïve Bayes method may be utilized to categorize the stride length for walking training.