Digit Recognition in Sign Language Based on Convolutional Neural Network and Support Vector Machine

Hearing and speech impaired individuals communicate using sign language among themselves and to the normal people. But it is challenging for non-sign language speaker to understand the sign language. Thus, sign language speaker or signer often finds difficulties in expressing their feelings. Hence, in this paper we have presented a method by combining convolutional neural network (CNN) and support vector machine (SVM) to recognize digits in sign language. We have employed the proposed model with two renewable Sign Language dataset namely MU_Handlmages_ASL and standard databases-SLD in order to measure efficacy of the model. The proposed model achieved average test accuracy 98.20% on ASL data and 98.30% on SLD data that indicates the model is very dependable and convincing to aid to communicate between signers and non-signers.