A vital element of technology for the deaf and dumb that can improve accessibility and communication is sign language recognition. The deaf and hard of hearing can engage with the hearing community thanks to SLR systems that can translate sign-language gestures into text. In this paper, an SLR system is developed using Convolutional-Neural-Networks (CNNs). The CNN-based SLR system undergoes several steps, such as data collection and pre-processing, data augmentation, model architecture design, model training, and evaluation and testing. The dataset of sign language gestures needs to include a wide range of motions and variants for the model to be guaranteed to generalize to new gestures. The purpose of data augmentation is to increase the model’s robustness and reduce overfitting, whereas data preparation includes cleaning, frame extraction, and normalization. A few of the layers that make up the model architecture are convolutional layers, pooling layers, and fully linked layers. Convolutional layers collect features from the input frames, whereas pooling layers reduce the spatial dimensions of the feature maps and add invariance to minute translations. The features are organized into different sign language movements by the layers that are completely integrated. The model’s performance can be judged using metrics like accuracy, precision, recall, and F1 score. People who are deaf or hard of hearing may benefit from SLR systems for better accessibility and communication, and CNNs have shown encouraging results in this area. The techniques outlined in this study can be used to develop an SLR system using CNNs and bridge the communication gap between the deaf and hard-of-hearing community and the hearing population.