Deep Neural Network based Sign Language Detection

Abstract

Deaf and dumb persons who are physically impaired use sign language to communicate. The main obstacles that have prevented much ASL study have been incorporated characteristics and local dialect variance in this work sets. To communicate with them, sign language should be learned. Peer groups are typically where learning happens. There aren't many study resources accessible for learning signs. The process of learning sign language is therefore a very challenging undertaking. Finger spelling is the first stage of sign learning, and it is also used when the signer is unfamiliar of the equivalent sign or when there isn't one. The majority of the currently available sign language learning systems rely on expensive external sensors. By gathering a dataset and using various feature extraction approaches to extract relevant data, this research discipline has been further advanced. The data is then entered into various supervised learning algorithms. The reason why the proposed results differ from existing research work is that in the developed fourfold cross validation, the validation set corresponds to the images of a person, which are different from the people present in the training set. Currently, the fourfold cross validated results are provided for various techniques.