Abstract
Ergonomic office setups are often used to prevent chronic pain caused by sitting at a desk for long periods of time. However, these setups are only effective if used correctly. This paper explores a solution that identifies poor posture in the three areas of the body. More ergonomic work office strategies include movements, such as transitioning from sitting to standing and vice versa. A user of this kind of system may need to monitor their posture for different setups. Previous studies have looked at monitoring human activities and static posture. However, most research focuses on one typical setup. The solution in this paper can be applied to three different ergonomic setups - a standing desk, a kneeling chair, and a typical office chair. The system consists of a wearable device, a stationary computer, and a simple user interface. The wearable device has accelerometers that monitor posture points on the upper legs, the torso, the shoulders, and the head. The data is sent wireless to be processed using a machine learning decision tree model. The result of the processed data is shown to the user in real-time. The result corresponds with a picture that helps the user visualize the posture issue detected. Five decision tree models were developed with at least 92 % accuracy. For each ergonomic setup, for each of the three sections of the body monitored, a model would be needed to be developed, resulting in nine different models. However, some of the body positions share similar posture positions across the three ergonomic setups. Reducing the number of models required less training data to be collected. This reduction did not seem to impact the accuracy of the models negatively.