Advancements in Hand Gesture and Sign Language Recognition Using BiLSTM- SoftMax and Deep-Learning

Gesture recognition, a crucial aspect of human-computer interaction, serves various purposes, from conveying emotions to facilitating communication between individuals and machines. Hand gesture recognition, particularly, poses a significant challenge due to its high-dimensional nature, making it an area of keen interest for researchers in pattern recognition. Addressing this complexity, this study explores different approaches to hand gesture recognition, emphasizing feature selection and extraction techniques to enhance machine learning model performance. Evaluating both manual and automatic feature extraction methods, statistical functions of central tendency were employed for manual extraction, while convolutional neural networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) models were utilized for automatic extraction. These features were then assessed using classifiers such as Softmax, Artificial Neural Networks (ANN), and Support Vector Machines (SVM). Among the models tested, the BiLSTM-ANN combination exhibited superior performance, achieving an accuracy of 99.9912%. This research contributes to advancing the field of gesture recognition, providing valuable insights for academics and professionals engaged in automated hand gesture detection and sign language recognition.