Prediction of “bad postures” based on Machine Learning models

The advances of the Internet, mobile communications and computing have opened new frontiers for a society that is increasingly focused on data analysis. New applications increasingly rely on machine-to-machine communications, which in turn create workloads that demand innovative systems, as network devices, sensors, and intelligent systems now generate vast amounts of data that can be processed to solve everyday problems, there is no doubt that new technologies have made our lives easier. However, despite all the advantages they bring, they have also brought some discomfort to our health. Especially, behind our back, which can be harmful to the spine. For these reasons, the purpose of this work was to design and build a prototype system that “detects and corrects bad posture” of a person when performing their daily activities. To do this, we use a prediction system with the help of three inertial sensors. The developed device consists of a monitoring system that includes 3 risk posture factors and is capable of providing feedback, through an application that alerts the user in order to prevent poor posture or inadequate posture that can become chronic. , causing neck pain, back pain, etc. In this way, a message is sent in case the posture of the human subject is outside the established threshold. The experimental results carried out with various machine learning algorithms demonstrate the abilities of the system to distinguish an improper head posture from a correct one. Different algorithms were applied, finding that by applying Neural Networks Artificial (ANN) we obtain better results.