Gesture Recognition Methods Using Sensors Integrated into Smartwatches: Results of a Systematic Literature Review
Pedro Raphael Inácio Gomes, Murillo Castro, Thamer Horbylon Nascimento
Abstract
This work addresses the importance of gesture recognition in smartwatches and aims to conduct a systematic literature review to identify the most commonly used methods and sensors in this field. The selected works were obtained from renowned databases, such as ACM Digital Library, IEEE Xplore, Scopus, and ScienceDirect. Initially, 265 articles were identified, and after applying inclusion and quality criteria, 43 articles were selected for detailed analysis. The systematic review allowed for the consolidation of existing knowledge and the identification of the most relevant approaches in this area. The results revealed a diversity of algorithms employed, such as Naïve Bayes, Artificial Neural Network, Support Vector Machine, Dynamic Time Warping, Symbolic Aggregate approXimation, Long Short-Term Memory, Random Forest, Multiple Dimensional Dynamic Time Warping, Gated Recurrent Unit, K-nearest Neighbors, Hidden Markov Model, Latent Dirichlet Allocation, and Convolutional Neural Network. Additionally, we found that accelerometers were the most commonly used sensors, followed by gyroscopes, magnetometers, barometers, vital sensors (EMG, FMG, PPG), and infrared sensors. The findings highlight the relevance of gesture recognition in smartwatches and provide valuable insights for researchers and professionals interested in this constantly evolving field.