Recognition of Sign Language Using Hybrid CNN–RNN Model

In India, the deaf community often utilizes Indian Sign Language (ISL) for communication. However, the absence of standardization makes it difficult to correctly identify and comprehend ISL gestures which obstructs efficient communication and accessibility. It is acknowledged that a strong recognition model that enables quicker and more accurate interpretation is necessary given the rising prominence of ISL as a key form of communication. In order to address this demand, we put forth a hybrid model for ISL recognition that incorporates Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). Our model makes use of an image dataset of 36 different ISL sign classes. In the beginning, spatial characteristics are retrieved from the input images using CNN, allowing the model to capture stated visual details. RNN is then used to capture the temporal connections within the sign sequences, improving the model’s capacity to identify complicated gestures. The results show that the hybrid CNN–RNN model gave higher accuracy of 98.2% in recognizing and interpreting ISL.