Source code in Linux for converting wave (or other audio) files into text!!

[antonela]

Hi!

I am desperately trying to find a source code in C, Fortran, Matlab or Visual Basic... (or anything else!!) which will be take an audio file as an input and will give a text file (ASCII perhaps) as an output! I run my thesis on Linux platform, so Windows-like programs don't work for me!

Anybody have ANY idea???

PLEASE HELP!!

[Force Flow]

Welcome to PCMech, antonela

Are you trying to make a transcript out of the audio file? Or do you just want to see the raw data?

[antonela]

Hi "Force Flow"! Thanks for your reply!

Actually, I don't quite understand what the "raw data" is... What I want is to convert an audio file to a file of type .txt (or any other text file) which will be "read" from a program (in Matlab) as input data.

Look, as fas as I know, Matlab can only read matrixes (numbers). So, what i really want is a way to convert "spoken words", such as "move" or "stop", into text.

For example: 0001 can stand for "move" whereas 0010 can stand for "stop"...etc...! Is this clear...? I hope you understand what I am trying to explain to you!!

If this is not possible... or if you don't know anything about this... any other suggestion is more than welcome!!!! Thanks again...

[doctorgonzo]

Why exactly do you need to do this? You are asking a lot, honestly. What you need is some kind of voice recognition software that will transcribe what is being said. Dragon NaturallySpeaking is an example of voice recognition software. But given how hard it can be to get voice recognition software working properly, it may be far easier in the long run to do the transcription yourself.

[antonela]

OK :-)

What I plan to do is a transcription myself! I don't want a software like Dragon Naturally-Speaking. What I can't figure out at all though is how exactly I can convert the audio file in a text file or anything else that can be read from another source code (preferably in C)! Or is this not necessary?

[Force Flow]

There's a program called WriteOutLoud which may do what you're trying to accomplish.

[GreenBMP]

I think he wants like the 0s and 1s in computer language of the file... that's the source in an audio-file right? Nothing is written in words.

[Force Flow]

He's trying to make a transcription of spoken words.

Once you have transposed your words, you can easly copy them into a txt file, which can be read by almost any independent language.

[antonela]

To start with... thanks everybody!!

Force Flow, this program you recommend unfortunately doesn't work for me. It runs on Window, and there is no source code... Yes, I want to make a transcription... But a file with 0s and 1s is just fine (actually is perferct!!) for me. Then, I will be able to correspond each set of 0s and 1s to a specific word.

Is that right or am I too optimistic????

GreenBMP, do you know I can get the source of a wav (or other) file? Or any other way to do that??

Thanks again guys!!

[doctorgonzo]

Originally posted by antonela Is that right or am I too optimistic???? Thanks again guys!!

I think you're too optimistic, unfortunately. The problem with transcribing audio to text is that there are a lot of different patterns of 0s and 1s that map to the same word; the human voice just isn't that exact, and so if you say the same word twice it is going to be coded to different bits. In addition, two different words can map to very similar bit patterns, making it hard to determine what word was spoken.

Here's a website with some speech recognition software for Linux. It may help.