Setup and Application of a Decision Tree Machine Learning in a Real Human Motion Tracking Scenario

Human motion analysis is becoming an increasingly used technology nowadays with the invention of smarter sensor technologies and with its applications in the field of health, ergonomics, sports, medicine, and so on. Turingsense EU Lab is one of the many companies around the globe working on this technology. The purpose of this thesis was to evaluate the SensorTile.Box PRO kit as a possible replacement of the standard Accelerometer/Gyroscope sensors that they use. The purpose of this possible replacement is the benefits that this box kit offers, allowing us to implement an A.I. based machine learning algorithm that could save both time and effort in the prediction of correct arm movements. The approach to investigate this evaluation was based on evaluating the performance of the box kit on predicting certain arm movements. This approach was based on first data collection, then training the decision-tree based ML algorithm by providing that data with their labels, and then finally testing the resulting configuration. By going through this process and modifying certain things along the way based on certain outcomes, the result was that the box kit was able to distinguish between the two different types of arm movements when it was attached to the wrist, and the accelerometer and the gyroscope data was used as the training data. Conclusions that were drawn from the series of these experiments were that the SensorTile.Box PRO kit is no doubt a very promising device with a lot of potential for applications in the field of movement analysis. However, particularly in this application for Turingsense EU Lab, its potentials are kind of limited as we do not have complete freedom in terms of what kind of training data we want to use, or what kind of ML algorithm we want to use. If these and other related issues can be resolved, then this box kit can come very handy to Turingsense EU Lab.