Hand motion detection is a critical step in the automatic translation of sign language, which allows deaf people to communicate. This research pursued to investigate the difficulties in character categorization in Indian Sign Language (ISL). In the related discipline of Sign Language, a considerable of the study was done, and the same could be said for ISL. The absence of appropriate datasets, occluded characteristics, and language variability with neighborhood have all been major roadblocks, resulting in little ISL research. In this paper, we proposed ISL detection and recognition using a deep learning framework. The Convolutional Neural Network (CNN) has been used for detecting the activity using the RESNET100 deep learning framework. The feature extraction and selection have been done in each convolutional layer and optimization has been done in the pooling layer. Multiple activation functions have been utilized in dense layers for classification on real-time hand gesture datasets as well as MNIST datasets. An extensive experimental analysis proposed system results has been evaluated with various machine learning algorithms, that produces almost 5-7% higher accuracy with state-of-art methods.

Hand gesture, Sign language, Indian signs, Deep Learning, Convolutional Neural Network, Object detection, Object movement