Individuals with hearing impairments face significant communication challenges due to a lack of universally accessible sign language systems. This paper proposes a sign language alphabet recognition system to address this issue by enabling accurate recognition and translation of American and Arabic sign language alphabets. The proposed Bilingual Sign Language Alphabet Recognition (Bi-LSLAR) system based on comparative analysis was conducted using five deep learning techniques, Convolutional Neural Networks (CNN), MobileNet, ResNet50V2, Xception, and DenseNet121. The proposed system is evaluated using the two benchmark datasets Arabic Sign Language (ArSL) and American Sign Language (ASL) Alphabet Datasets. The proposed system has an interactive interface that enables real-time sign language recognition. The interface is developed using OpenCV, the world's largest library for image processing and computer vision applications. This interface captures input from videos, recognizes gestures from the hands, and converts them into text that can be seen in either Arabic or American. MobileNet is the most efficient technique in the proposed system, with accuracies of 97.17% for ArSL and 98.97% for ASL. These results highlight the system's potential for enhancing accessibility and real-time sign language translation, hence facilitating communication between hearing and hearing-impaired individuals.