Bangla Sign Language Recognition from Hand Gestures using Convolutional Neural Network

In modern society, sign language has become the primary language of the deaf and dumb community. To communicate with others, they employ a variety of signs. Sign language recognition is a new field of study that aims to improve communication with the deaf and dumb. Many studies on the identification of Bangla sign language have been reported in the literature. However, we found very little research with high accuracy for the identification of Bangla sign language. In Bangladesh, there is also a sizable deaf and dumb population. In this research, we build a modified convolutional neural network for recognizing Bangla sign language. We have 37970 data for 59 classes in our dataset. Finally, we looked at our model’s performance separately. We got 100% accuracy for digits, 99.84% for alphabets, and 99.5% for combined use.