Performance comparison of machine learning and neural network for sitting posture detection among university students

Sitting occupies more hours of the day especially for typical adults due to the global Covid-19 pandemic. Research had shown that prolonged sitting is unhealthy, which may lead to health concerns such as musculoskeletal pain. Therefore, sitting adults should aware and notified if there is any improper sitting posture. This work was conducted to detect a person's current spinal position over a period of time using tri-axial accelerometers. The data collected from sensors were processed and classified using Machine Learning (ML) and Neural Network (NN) classifiers. By extracting axial displacements between sensors, spinal movements can be determined. Data was collected from a total of fourteen university students using the constructed sensing circuitry and were transferred to Matlab for processing. The results achieved accuracies of 93.7% and 96.7% for ML and NN, respectively.