A Sitting Posture Monitoring System in Wheelchair Users

Postural monitoring is essential to prevent musculoskeletal problems and to carry out functional tracking of wheelchair users. Currently, this monitoring is carried out by means of eventual health observation or questionnaires that carry an implicit subjective character. In order to provide health care specialists with objective and quantitative information on the postural status of patients, in this chapter a sitting posture recognition intelligent system is presented. The novel monitoring device, based on pressure sensors, allows the recognition of common sitting postures in wheelchair users. It has been designed determining the optimal sensors location, reducing costs while maintaining information significance. With this information as input, different machine learning techniques are analyzed for posture classification, based on a stratified K-fold methodology. In addition, an analysis of the optimal number of sensors to obtain a high percentage of success has been carried out, using Random Forest. The results show that the approach used is suitable for posture recognition, with accuracy over 90%, and that a number of nine sensors are enough to distinguish between different postures. In this way, we are able to recognize sitting postures with as few sensors as possible and with the lowest possible computational cost.