ABSTRACT
The whole concept of technology and research is based on making the existing technologies more efficient and accurate by overcoming the challenges that we are facing in the respective field. It's a never-ending process. One such challenge is to make communication easier for deaf and dumb people. Luckily, sign languages have come to the rescue. Sign language is the communicating mode for deaf-mute. Normal people may not be able to understand sign language to communicate with deaf-mute people and this problem needs to get overcome. There have been some advancements in this field in recent times as well as some applications have been developed for the detection of sign language and many research papers have also been published in the same field. However, one challenge is followed by other challenges and now the race is after coming up with methods more accurate than the already existing ones to communicate with the deaf and dumb people. This chapter proposes a LeNet-like architecture for classification to get high accuracy. We choose the Modified National Institute of Standards and Technology (MNIST) database sign language data set to train and test our model. Our proposed model has given an accuracy of 99.60% that is the admissible accuracy. And this chapter will also deal with the over fitting issues with the regularization method.
