Attention Trinity Net and DenseNet Fusion: Revolutionizing American Sign Language Recognition for Inclusive Communication

American Sign Language (ASL) is a unique and essential language for the North American deaf and hard-of-hearing community, with distinct grammar and expressive hand and facial movements. In this study, we embarked on a unique journey to enhance ASL recognition, recognizing the significance of this language as a means of communication. Our approach began by consolidating two distinct datasets, amalgamating their valuable images to create a comprehensive corpus for classification. This fusion of resources provided a robust foundation for our research. A hallmark of our methodology is the incorporation of three sophisticated attention modules: channel attention, soft attention, and squeeze and excitation attention. These modules collectively form the framework known as Attention Trinity Net (ATN), a testament to our commitment to precision and performance. Furthermore, we introduced the concept of Uni Attention Network (UAN), an innovative approach that leverages DenseNet121, a powerful neural network architecture. Each attention module, channel attention, soft attention, and squeeze and excitation attention, is thoughtfully integrated into the UAN framework, enriching its ability to discern intricate ASL gestures. In a climactic juncture of our study, we seamlessly combined the outputs of the three UANs based on the ATN to form a cohesive and comprehensive prediction model. The synergy achieved through this integration resulted in an astounding accuracy rate of 99.98%, surpassing the performance of any existing ASL recognition models. Our study is a significant milestone in ASL recognition, highlighting the potential of advanced attention mechanisms and unified network structures in deep learning and computer vision. This achievement benefits the deaf and hard-of-hearing community and advances artificial intelligence and human-computer interaction.