Isolated Sign Language Recognition Using Deep Learning

Communicating with individuals who have hearing disabilities presents a significant challenge. Deaf and mute individuals rely on sign language to communicate, making it difficult for hearing individuals to understand. To bridge this communication gap and promote social inclusion, we focus on developing a system to recognize sign language gestures and convey information to hearing individuals. We conducted extensive research in this field by reviewing relevant research papers. Throughout the project, we explored various models and datasets, explicitly detecting word-level sign language using Long-term Recurrent Convolutional Networks (LRCN), Convolutional Neural Networks (CNN), and Convolutional Long Short-Term Memory (ConvLSTM) models on the WLASL dataset. Our approach involved multiple steps, starting with extracting frames from video data. We then applied preprocessing techniques to enhance the quality of input images, leading to improved model performance. Additionally, we utilized image segmentation methods to isolate the hand from the background, enabling our models to focus on relevant regions and enhance detection accuracy. In the future, we aim to incorporate word-level annotations to facilitate sentence-level and real-time sign language detection.