Classification of Lying Postures

The aim of this paper is to classify participants’ lying postures using publicly available data obtained from matrix pressure sensors. Two machine learning algorithms, Convolutional Neural Network (CNN) and Random Forest (RF), are employed for subjects’ postures classification. The evaluation revealed that CNN applied on pressure topographic maps (accuracy of 90%) outperformed the RF algorithm that utilized extracted features (accuracy of 64%). Moreover, additional transformations of topographic maps improved CNN accuracy up to 96%, indicating that processing may play a significant role in the application of deep learning algorithm for lying posture classification.