Hand gesture recognition has become an indispensable part of human-computer interaction. It supports intuitive, contactless, and accessible means of communication. This work proposes a robust deep learning-based system for the real-time static recognition of hand gestures to assist communication with the deaf and hard-of-hearing. We propose hybrid architecture that consists of the Swin Transformer for contextual attention, ResNet34 for spatial feature extraction, and a BiLSTM layer for temporal understanding. The model was trained and evaluated using a subset of the HaGRID dataset, consisting of 18 different classes of hand gestures. It achieves a training accuracy rate of 99.3% and a testing accuracy rate of 98.03%, thereby demonstrating its excellence in performance amid different lighting and backgrounds. Our approach proposes a scalable solution that can be integrated into assistive technology for gesture-to-text or gesture-to-speech conversion, thus bridging the communication gap for deaf individuals.