SPEECH TO TEXT CONVERTER FOR PARTIALLY DEAF PEOPLE
Aspiration behind this innovative idea came from the problem that our young people are facing i.e. partially deafness. Our study leaded us to fact that there are so many people in Mumbai region facing partial deafness as those others all around the globe. So we try to come up with a unique engineering solution “C-YANTRA.”
C-YANTRA is designed for an interactive classroom of partially deaf people. There are many devices in market for deaf people like cochlea implant, CART (Communication Assess real time transition) etc. which gives tremendous economic stress as well technical issues for many people. In CART system there is restriction on no of speakers as well teacher should always face towards students which is obviously not possible in real lie. So we are introducing a new idea here with ‘C-YANTRA’ It’s a like a close friend of user, highly efficient and interactive GUI which makes it worth. It can be designed in both formats as in wearable as well as one time installation format. No internet connection is required but if it is der then it can connect the user to the distant peoples.
‘C-YANTRA’ is a purely signal processing and simulation system with well featured GUI in the form of a common screen (In classroom demonstration we can use our blackboard as a screen)each person could be differentiated by his/her fundamental frequencies. Everyone has different vocal track/velum structure, thereby different pitch, tonality, loudness, sharpness etc.
Hardware and simulation part of our system...
For prototyping we used CRIO 9075 as a FPGA and C 9233 as analog input module for acquiring speech signals by National Instruments. This is completely an autonomous system. After the deployment of the algorithm of system teacher and student have to give the introduction to the system (they need to speak in front of the system where microphones are fixed to record the samples of user)this information will be stored into the system. Once class begins the system will decode the sound of individual from the mixture using ICA (Independent Component Analysis) and display the words/text corresponding to the sound or voice using(HMM Hidden Markov Model) on to the screen along with gender, name, emotion of the respective speaker. In this way we can overcome the barrier between the partially deaf and normal once. Prototype is still under development.
Product realization and market analysis:
This idea can be implemented on a mass development basis. Any leading technical firm can take this idea to the society... Strictly talking in terms of market conditions for this product it would maximally cost as just buying another iPad on mass production scale. Any tablet manufacturing company would easily manufacture such a product using wearable electronics components with aim of Tech for EveryOne