Abstract
People commonly communicate in English, Arabic, and Bengali spoken languagesthrough various mediums. However, deaf and hard-of-hearing individualsprimarily use body language and sign language to express their needs andachieve independence. Sign language research is burgeoning to enhancecommunication with the deaf community. While many researchers have made stridesin recognizing sign languages such as French, British, Arabic, Turkish, andAmerican, there has been limited research on Bangla sign language (BdSL) withless-than-satisfactory results. One significant barrier has been the lack of acomprehensive Bangla sign language dataset. In our work, we introduced a newBdSL dataset comprising alphabets totaling 18,000 images, with each image being224x224 pixels in size. Our dataset encompasses 36 Bengali symbols, of which 30are consonants and the remaining six are vowels. Despite our datasetcontribution, many existing systems continue to grapple with achievinghigh-performance accuracy for BdSL. To address this, we devised a hybridConvolutional Neural Network (CNN) model, integrating multiple convolutionallayers, activation functions, dropout techniques, and LSTM layers. Uponevaluating our hybrid-CNN model with the newly created BdSL dataset, weachieved an accuracy rate of 97.92\%. We are confident that both our BdSLdataset and hybrid CNN model will be recognized as significant milestones inBdSL research.