Answers are grounded in this paper and cite their sources.
Hybrid Vision Transformers combined with Ensemble Transfer Learning improve ISL recognition accuracy, suggesting future applications in assistive technologies.
Indian Sign Language (ISL) identification methods play a central role in enhancing communication between hearing-impaired and non-impaired individuals within their community. However, modern ISL identification algorithms face challenges due to hand gesture variability, complex visual settings, and limited official annotations. This study proposes a Hybrid Vision Transformer with Convolutions (HVTC) combined with Ensemble Transfer Learning (ETL), incorporating advanced transfer learning methods such as Adaptive Lightweight DenseNet, VGG19, and XceptionNet for Multi-Task Learning, along with ResNet with Dynamic Depth and MobileNetV3 with Attention Mechanisms to improve ISL recognition accuracy. Four primary challenges affect ISL recognition: obstructions in the camera view, inconsistent lighting conditions, visually similar motions that are difficult to distinguish, and the need for extensive labeled datasets for deep learning systems. The ETL-HVTC processing method effectively extracts spatial-temporal motion data by leveraging sophisticated neural network algorithms. Transfer learning reduces dependency on large datasets, while the ensemble approach integrates multiple predictive models to enhance model stability. A robust ISL recognition algorithm should prioritize real-time capabilities, high recognition accuracy, and an expanded application scope. Secure gesture dataset pre-processing enables the optimization of hybrid ViT Large Model-CNN models, where collaborative learning ensures reliable classification outcomes. Experimental results demonstrate that the proposed ETL-HVTC system outperforms independent ViT Large Model and existing CNN models on ISL benchmark databases in terms of precision, recall, F1-score, and accuracy. The implementation approach yields fast and effective results, facilitating the development of assistive devices that promote more inclusive communication for individuals with hearing impairments.
No takes yet. Share an insight, caveat, or question.
Anand et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: