Sign Language Recognition (SLR) has recently emerged as a crucial technology for facilitating communication between the hearing and deaf communities owing to the advancements in Machine Learning (ML). This paper surveys recent developments in SLR, focusing on widely appreciated and adopted techniques like Convolutional Neural Networks (CNNs), Transformers and Long Short-Term Memory (LSTM) networks. Frameworks such as MediaPipe detect hand, body, and facial landmarks with precision. Through comprehensive analysis and comparison of various approaches based on performance metrics including accuracy and Bilingual Evaluation Understudy (BLEU) scores, we identify the strengths of different architectures across static and dynamic gesture recognition techniques. Analysis reveals that MediaPipe-based frameworks combined with LSTM temporal modeling show promise in real-time applications. Building on previous performance and results, a real-time SLR system is presented that utilizes positional shifts in selected landmarks generated by MediaPipe to model gesture dynamics, alongside LSTM for temporal sequence modeling. The trained model then predicts gestures from live-fed video, ensuring efficient and accurate SLR.