Neural speech — An aid for differently challenged

Abstract

The world today has growing physically challenged people especially those who have hearing-speaking impairment they face several problems in communicating with people around them and also with others around their circle as they need to be familiar with their sign-language. Current technological solutions for this include voice to word/text generator, glove based (super-glove) for hand gesture image capturing. But, these methods/techniques have disadvantages especially in the case of voice to word-/text generation it becomes easy for the common people to communicate with the differently challenged but the reverse requires knowledge of sign-language which everyone will not be familiar with, in the case of glove based technique it requires a specially designed glove to be worn by the differently able this becomes an additional hardware requirement, moreover there is a delay between input and output too. In order to overcome such drawbacks Here we have brought in a solution using Dynamic Bayesian neural systems which has efficient generalization ability, tolerance to input noise, Can handle high dimensional output easily, high speed, parallel processing and mathematical modeling is not required.