USE OF INERTIAL MEASUREMENT UNIT IN CLASSIFYINGWALKING AND RUNNING MOTION WITH MACHINELEARNING METHOD

• Main Author: Khairunnisa, Juanet
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74923
• Institution: Institut Teknologi Bandung

At the Kuala Lumpur SEA Games in 2017, there was controversy in the women's 10,000- meter sprint race involving the winner, Elena Goh Ling Yin of Malaysia. Vietnamese media accused Elena of cheating by running instead of walking according to the race rules. Manual assessment of footwork by judges is subjective and prone to error. Therefore, the development of an objective system is required. In this research, an IMU sensor is used to measure and record the athlete's body kinematics data with high accuracy, so it is hoped that a system can be developed that is able to objectively analyze and classify the athlete's footwork. This research focuses on developing a simple IMU platform system that can analyze motion cycles and classify walking and running movements in humans. Wireless IMU sensors will be placed on the heel and metatarsal to collect acceleration and angular velocity data. These data will undergo Butterworth filtering and Min-Max normalization processes. Each motion cycle will be used as training and testing data for various machine learning algorithms. The main objective is to compare the accuracy of different algorithms and select the one with the highest accuracy. The research results show that using the Random Forest algorithm with two IMUs achieves an accuracy of 86.20% in recognizing walking and running movements, with an execution time of 1 second. Angular velocity in the y-axis of the metatarsal sensor proves to be a significant kinematic parameter in classifying movements. Additionally, additional information such as Confusion Matrix, Feature Importance Score, and Shapley Additive Explanation are employed to analyze the influence of 12 observed kinematic parameters