Sign language and hand gestures are vital forms of communication for the deaf and hearing communities. Assistive technologies are being developed to translate sign language movements into text to close the communication gap. This paper compares two deep learning architectures, LeNet and AlexNet, for sign language gesture recognition and text conversion using ANN and CNN. Our proposed research focuses on evaluating the performance of LeNet and AlexNet on a comprehensive dataset of sign language gestures, with a specific emphasis on accuracy. The results indicate that while both architectures show promising results, LeNet is better and more efficient for translating sign language gestures into text when you have limited computational resources. This work advances assistive technology by shedding light on the efficacy of deep learning architectures for sign language communication. Better communication tools for those who are hard of hearing may result from this research. Lenet has achieved a 97% accuracy and AlexNet has achieved 27% accuracy in recognizing hand gestures using depth images. Moreover, the researchers have tested the models on various sign language datasets to assess their generalizability and have evaluated their resilience to variations in lighting conditions, hand orientation, and signer diversity.