Enhancing Real Time Hand Gesture Recognition Through Ensemble Voting Method

Human-Computer Interaction focuses on the interfaces between humans and computers and the ways in which they interact. In this project, we propose a real-time hand gesture recognition ensemble model that integrates multiple Deep learning algorithms such as Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) along with Unsupervised Machine Learning algorithms including k-Nearest Neighbors (kNN), Support Vector Machine (SVM. Our aim is to develop a robust system capable of accurately interpreting hand gestures in real-time with high accuracy and low recognition times, under various environmental and lighting conditions, to facilitate intuitive human-computer interaction. Through an ensemble learning technique such as majority voting, the prediction of these individual classifiers was combined to enhance performance to predict the accuracy of recognition and average recognition time was calculated. The resulting ensemble model was implemented in a real-time system by capturing live video input from a camera. The following accuracy values were obtained in the Best-Case scenario, 0.988 for KNN, 0.964 for SVM, 0.94 for LSTM, and 0.97 for GRU. The study improves gesture-based interfaces and offers developers and researchers valuable information.