 This paper presents a new approach to postural monitoring of wheelchair users. It uses a multi-layer neural network to classify different postures from sensor data collected by a novel monitoring device. The model is trained using a stratified k-fold cross-validation technique, which allows it to generalize better across different subjects. This makes it suitable for use in both familiar and unfamiliar situations. The system can then be used to help wheelchair users and healthcare professionals monitor their posture, regardless of physical complexity. This article was authored by Patrick Vermander, 8th Cyberman Cicidor, Itseer Cabanes, and others.