Sign Language Recognition for Deaf People Using Deep Learning

Recent deep learning breakthroughs will make sign language easier to decipher, which will be vital to enable deaf individuals to communicate on a daily basis. Deep learning will be utilized in this study to convert sign language gestures into text, making it easier for everyone to know what deaf people are communicating. To enable this to function well, video data of sign language will be gathered using a software named OpenCV and will be thoroughly prepared for analysis. A unique type of neural network named a 3D Convolutional Neural Network (3D-CNN) will be constructed which can identify signs created using a single hand. In this project, various methods will be experimented with to train the neural network, including with optimizers such as Adam and Stochastic Gradient Descent (SGD). These optimizers will assist the network to learn quicker and better from the gathered data. Adam will tune the rate at which the network learns for every section of the data, which will be fantastic for large datasets. SGD will perform well when the data is not very distinguishable or is missing. Once we have done all of that, our system will be capable of translating sign language movements into text with a high degree of accuracy. This will result in it accurately understanding and interpreting signs into words the majority of the time. This technology will be significant because it will facilitate bridging the communication gap between sign language users and all other people. In the future, we aspire to further improve our system so it can assist more individuals in real-time conversations.