How do i develop speech recognition software in c++?

HI


I want to start working on final year project (computer science)


I think about speech recognition - something like converting sound of the lector to words in microsoft word(good for students). how can I implement such project (where to start?).

How do i develop speech recognition software in c++?
Hello,





Microsoft has already released a Speech Recognition framework for you to use. It is called SAPI, Speech API.





"Microsoft Speech" http://www.microsoft.com/speech/


Where the newest SDK is called 'Microsoft Speech API 5.3'





For example, I have done a simple C# application that when I say "Volume Higher" it increases volume of the PC. I have interop the C++ Speech SDK libraries. Now I realized Microsoft created a Managed API which would make it even easier





It is easy to use, since your in Computer Science, it will be easy for you to understand the SDK in C++. Search for SAPI Tutorials on Google, and you will find alot of tutorials :) I bet you can setup a simple speech application in an hour or two.





Good Luck





======


Edit: Update:





Man, many people here make it seem so difficult to create speech recognition software. Dont' fall for that rumour, creating Speech Recognition is SO EASY if your using Speech SDK SAPI.





Sure creating a Speech Recognition engine is a lot of work, but why reinvent the wheel if Microsoft has a framework that does that for you? But since your trying to copy the voice from the Lecturer to Text, then you have to apply some filtering to that voice so it could be clear.





In University, for one of my labs, I need to control a robot, my teammates and I were bored that hour, so we applied the SAPI to the robot and with simple speech to text, we said MOVE, LEFT, RIGHT, MOVE, LEFT LEFT LEFT. Pretty straight forward.





That SDK allows you to train as well, so you can train your keywords.
Reply:We have a guy working at the company where I work programming speech recognition software. That's all he does, he's good at it, he's been working on the module for 2 years. It is very difficult to get it right, and in the end it's always fallible if you find someone with a thick enough accent.





I think you should consider a different final year project if you hope to complete it.
Reply:if you wanna work on speach recognition, then first of all you should know about neural networks, feature extraction and....


and after that you should try to implement those algorithms in c++ or any other language you want. why dont you try MATLAB? it has many toolboxes for signal proccessing, you can write codes easiely in .m files and then convert them to c or c++ codes using MATLAB compiler...
Reply:First you need to decide on what the scope of the project is.





Are you going to implement the speech recognition engine or simply build an application that does speech recognition?





If you are to implement your own speech recognition engine, then you might want to learn about machine learning, signal processing, collect a lot of voice samples, and build a system to train your engine.





If you want to simply implement an application that types what is dictated, then you can use a speech recognition library. Check out http://tldp.org/HOWTO/Speech-Recognition... to see the various available libraries.





Once you decode the incoming stream into text, you will need to build a module to format the output according to the different formats you want to support, or a module to integrate with a text editor so it would type what is said in real time.





There are other modules to worry about to support corrections and such, but you should start with a small thing first.
Reply:It would be easier if you did audio compression and or processing.





Good luck getting a speech recognition to work in a final year project.