FC-YOLO: A High-Precision Sign Language Recognition Algorithm Based on Global and Local Feature Collaborative Learning

Sign language recognition is a crucial communication tool for individuals who are deaf or hard of hearing, with its accuracy and efficiency directly impacting their daily interactions. However, existing sign language recognition models often rely heavily on global features while paying insufficient attention to subtle local details. This can decrease recognition accuracy in practical applications, ultimately hindering effective communication. To address this issue, we propose FC-YOLO, a sign language recognition algorithm that integrates collaborative learning of both global and local features. In this network, Frequency Adaptive Dilated Convolution (FADC) is employed to dynamically adjust the receptive field, effectively capturing high-frequency information and significantly enhancing sensitivity to local details. Additionally, a Cross-layer Convolution and Attention Mechanism Feature Fusion Module (CAFM) is embedded, enabling complementary integration of global semantic information and regional features, thus providing a solid foundation for recognizing complex sign language symbols. On the American Sign Language (ASL) dataset, this method achieves 97.2% accuracy, 93.3% precision, and 99.5% mean average precision.