Problem Statement

- It's a speaking and listening Calculator for Blind people.
- It will take the digits and operator from the user using voice input.
- Output will be the calculated result in audio format i.e. speech .
- Right now I had broken the numbers into digits i.e. 99 will be feeded as 9 followed by 9
- For now,I had restricted the operators to additon,substraction,division and multiplication.

Design Overview

User Input(Speech to Text)

For this I will be using cmusphinx API,which is an opensource from CMU.

I have to define the grammar for the words to be recognised in the separate file.(my part)

Pronounciation of the words specified in grammar will also be present in separate file(as per convention of CMUSPHINX)(my part).

I will be using Multi-Threading so to take speech input,do calculation on text and convert speech back to text(my part).

Text-to-Speech

API that I will use is,TTSLIB,which is built on top of 'freetts' API.

This is also an open source API.

Model

I may introduce one more module before processing and calculation module(My module)

for the Validation,Verification and Conformation of the input.

Implementation Plan

Tools/Technologies

Java
CMUSphinx(Speech to Text API)
TTSLIB(Text to Speech API)
Eclipse

    Amount of Code : Around 500 Lines.

    Time Needed : 1 week for exploring API'S and 2 weeks for building calculator on top of it.

    Current Status : 30% done.

Testing/Evaluation Plan

To test or evaluate project,following steps are need to be done.

The correctness of this project depends upon the reliability of the speech to text API.
Pronunciation of the words matter here.Pronunciation should match with specified one in the file.
Speech is independent of the type of the voice,so no need for different voice testing.
Text to speech synthesis reliability also depends upon the Library.
Big TestSuite is not needed for this project.

Context

Project has endless scope,The more you teach to computer for doing task,more you will get the work done.

My perception and anticipation for the project was to allow blind people to do (fun)programming.

The way programming evolved was through ABACUS,so I thought this as a starting point.

It can be extended for multi-lingual speech recognition and synthesis.

It can also be extended for doing various complex Operations like log,exponent,factorial etc.

References

CMUSPHINX(Speech to Text API).

TTSLIB(build on top of freetts).

Complete Reference JAVA ,Herbert Schildt.