Evaluation of Motion-Capture Suit Data and Gesture Recognition Using LSTM and GRU Neural Networks

New forms of human-machine interface (HMI) and human-robot interaction (HRI) are key challenges in information and communication technologies. This research explores gesturebased control using data from a motion-capture suit to enable interactions with mechatronic devices. The study achieved its objectives by designing, programming, and training two neural network models: LSTM (long short-term memory) and GRU (gated recurrent units). An application was developed for performance testing and visualization, utilizing data from the Rokoko Smartsuit Pro II. Results demonstrated the effectiveness of these neural networks and provided a detailed analysis of the technology used, showing strong potential for practical applications in HMI. The outcomes align with the trends in Industry 4.0 and the emerging human-centric Industry 5.0, where gesture control could enhance ergonomics in industrial and service workflows. The results will be useful in further research where this method may be directly applicable to mechatronic devices.