Sign Language Recognition using Transfer Learning

For the community of hearing-impaired people, sign language recognition is essential for bridging communication gaps. This paper provides a novel approach that combines transfer learning with the VGG-19 convolutional neural network architecture to enhance the accuracy and efficiency of sign language recognition systems. The suggested method trains the network to recognize the distinct characteristics of sign language motions by utilizing the pretrained VGG-19 model on a sizable image dataset of 300 images for each sign which were fed into the VGG-19 model for training and testing purpose. The model was optimized for precise recognition of various sign language signs by fine-tuning. This reduces the requirement for vast labelled data and training time. This model was successfully implemented by converting the signs to their text equivalents and achieved 99.92% accuracy. Experimental results shows that this strategy is successful and achieves better performance than conventional methods.