Marker-Based Monitoring of Seated Spinal Posture Using a Calibrated Single-Variable Threshold Model

This work, as part of a larger project developing wearable posture monitors for the work environment, seeks to monitor and model seated posture during computer use. A non-wearable marker-based optoelectronic motion capture system was used to monitor seated posture for ten healthy subjects during a calibration exercise and a typing task. Machine learning techniques were used to select overall spinal sagittal flexion as the best indicator of posture from a set of marker and vector variables. Overall flexion data from the calibration exercise were used to define a threshold model designed to classify posture for each subject, which was then applied to the typing task data. Results of the model were analysed visually by qualified physiotherapists with experience in ergonomics and posture analysis to confirm the accuracy of the calibration. The calibration formula was found to be accurate on 100% subjects. This process will be used as a comparative measure in the evaluation of several wearable posture sensors, and to inform the design of the wearable system