Analysis of Lifting Posture by Two Inertial Measurement Units and a Classification Model Based on a Convolutional Neural Network

An improper lifting posture can lead to discomfort in the lower back. A more effective approach to preventing recurrent lower-back discomfort is to detect and assess proper lifting posture before an injury occurs, particularly because the recurrence rate remains high after treatment. To evaluate the lifting posture, we used two inertial measurement units, placed at the L5 and L1 spine locations, to capture posture data during lifting. Additionally, we proposed a classification model based on a convolutional neural network to classify the lower-back posture and kinetic chain. Our experimental setup included five participants with fitness experience. We collected data from these participants to train and validate the accuracy of our classification model. This model excels at categorizing the four types of lifting postures, achieving an overall classification success rate of 90%. This research paves the way for future applications that correct posture, which could be highly beneficial for gymgoers and individuals experiencing lower-back discomfort. By potentially providing valuable postural information, our aim is to assist these individuals in avoiding potential injuries.