Human body action recognition with machine learning for bionic applications: a Sensor Data Fusion Approach

Health sensors are a valuable source of data for researchers investigating the body’s varied activities. Human activity recognition systems have a special attachment to health sensors as their construction requires a focus on several steps such as pattern classification, feature extraction and selection, classifier design, and the learning process.Our research is primarily concerned with recognizing the various movements of the body’s extremities in order to contribute to research on prosthetic control employing electromyography signals and the development of smart health systems. As a beginning stage, we decided to focus our research on detecting physical activity by utilizing data from two sensors with the intention of improving analysis and hypotheses associated with bionic applications and artificial intelligence solutions for health systems.We demonstrated this approach with a specific case study on the exploitation of the data generated by the accelerometer and gyroscope. This analysis was performed on seven human activities along with a new set of indicator parameters, statistical features, and five well-known machine-learning classifiers to evaluate the results.