Human Upper Limb Motion Recognition Using IMU sensors and Artificial Neural Networks

Abstract

Computer vision researchers have been tasked with developing applications to recognize the human body's structure, movements, and tracking. Recognizing human upper limb movements has brought benefits in physical therapy, virtual reality, human-robot systems, sports. This paper presents a multi-layer perceptron (MLP) artificial neural network (ANN) designed to recognize the upper limb movements performance in daily life tasks. For this, inertial measurement units (IMU) were employed to acquire information regarding of the upper limb movement. The performance of the model was assessed through a confusion matrix and the receiver operator characteristic (ROC) curves. Recognition accuracy obtained for the ANN model was 97.39%, the mean area under the curve (AUC) of the ROC curves was 0.973. According to the results, the proposed ANN could recognize upper limb movements in daily life tasks. Further research is suggested to test the model with fine movements.