Indian Sign Recognition System using Adaptable Segmentation and Machine Learning

Deaf community all over the world uses sign language as a means of communication. Although it resolves the communication barrier among the community itself, the integration of acoustic and sign language is still matter of research. This paper proposes a dynamically adaptable image recognition module coupled with convolution neural network to recognize hand signs. The paper addresses the problem of isolation of hand from a robust environment and recognizing the hand sign. A series of transformation and segmentation is applied to the input, which in turns helps in feature extraction. Preprocessing of both training and input and is done to give a performance boost. Comparative study is drawn in the end to find the effect of pre-processing, of input and training data, on the classification speed.