Gesture Recognition in Smartwatches Using LSTM for Interaction in Low-Cost Virtual Environments
Leonardo Silva, Fabrizzio Alphonsus Alves de Melo Nunes Soares, Juliana Paula Félix, Luciana Cardoso, Renan Vinícius Aranha, Thamer Horbylon Nascimento · 2024
This work presents a methodology to enhance interaction in VR environments using accessible devices such as Google Cardboard and smartwatches. The main contribution is the implementation of a recurrent neural network model of the LSTM type, designed to recognize gestures captured by smartwatches. This enables more natural and fluid interaction with the virtual environment, significantly elevating the level of immersion and responsiveness perceived by users. Preliminary results demonstrate that the LSTM model achieves robust performance in accurately identifying gestures, which is essential for providing an immersive and engaging experience. We hope that this approach expands the possibilities for interaction in virtual environments and represents a significant advancement in the field of wearable computing applied to Virtual Reality.