Detecting hand movements is a crucial part of translating sign language, which helps deaf people communicate. However, research on Sign Language faces several challenges. Sign Language lacks appropriate datasets, and it has issues like obscured hand gestures and variations in how people use the language in different regions. These obstacles have hindered progress in sign language research. In our study, we proposed a solution for detecting and recognizing sign language using deep learning technology, specifically a Convolutional Neural Network (CNN) called RESNET100. We used this network to identify hand movements. In each layer of the network, we extracted and selected important features, and we optimized the network's performance in the pooling layer. To classify hand gestures in real-time, we used various activation functions in the dense layers and tested our system on real-time hand gesture data and the MNIST dataset. Our experimental analysis showed that our system achieved a significantly higher accuracy, about 5-7% better, compared to the best existing methods in the field. This means our approach has the potential to improve sign language translation and communication for deaf individuals.