Dynamic Sign Language detection system using Media Pipe Holistic and LSTM based Deep learning Model

A significant handicap is a loss of verbal communication. Sign language serves as one of the greatest popular ways for people with various disabilities to communicate, while they utilize a wide range of other methods as well. Sign language is the means by which people who are deaf or hard of hearing may communicate with one another and with the hearing world. This research describes a new architecture for deep learning by utilizing Media Pipe Holistic in conjunction with the long-short-term memory (LSTM) network. The LSTM network is employed for classification, while spectral and spatial data are extracted using Media Pipe holistic. 2,000 videos were used to train the model, with 80% used for instruction and 20% used for verification. On the training data, we got an accuracy of 93%, while on the testing data, we got 96%. Purpose of this approach is to develop a way for interaction with ordinary people and hearing challenged.