Abstract:

Abstract This paper addresses the urgent need for multilingual hand gesture detection that is accurate and efficient for use in real-time applications. The precise recognition of hand gestures across several languages becomes essential with the growing integration of gesture-based interfaces in a variety of applications, such as virtual reality, augmented reality, and human-computer interaction. When dealing with continuous gestures and many characters per gesture, present models struggle to achieve high precision, accuracy, and recall. The suggested solution employs a unique ensemble learning model that takes advantage of interconnected key point analysis of single hand motions to get around the shortcomings of the existing algorithms. A complete set of 441 distance characteristics is achieved by modeling hand motions using 20 key points, including 4 key points for each finger and one for the hand's center. Then, linguistic characters in Hindi, English, and Marathi are attached to these features. Using Ant Lion Optimization, which ensures high variance feature sets are used in the ensuing training of the ensemble learning model, the selection procedure is carried out in order to maintain the most discriminative and informative features. A variety of classifiers, including Naive Bayes (NB), k-Nearest Neighbors (kNN), Support Vector Machines (SVM), Multilayer Perceptron (MLP), and Logistic Regression (LR), are combined in the ensemble learning approach. This combination of classifiers strengthens the model's general robustness and increases its capacity to generalize successfully across many languages. Additionally, by incorporating a state machine that permits the seamless processing of continuous gestures, the model successfully handles characters that require numerous gestures. The suggested model's remarkable performance metrics and computational effectiveness are its main benefits. The results of the experiments show a remarkable precision of 99.8%, accuracy of 99.5%, and recall of 99.4% with delay lower than 2.5 ms for different use cases. These outcomes outperform the effectiveness of current approaches, making the suggested paradigm very effective for real-time scenarios and enabling fluid interaction with gesture-based interfaces. The paper concludes by presenting a novel ensemble learning model that, through interconnected key point analysis, efficiently recognizes multilingual hand motions. The model performs remarkably well because to the integration of many classifiers and Ant Lion Optimization feature selection, making it ideal for real-time applications and making a substantial addition to the study of gesture detection process.

Keywords: