make spoken words machine recognizable.

make spoken words machine recognizable.

[paul fellows]

A better way to explain why I want to add echo to a voice to make spoken words machine recognizable. 
Current voice to text systems use very complex mathematical algorithms to work out what are words and what are not words, and then to work out what the word might be.

-d gain -n -D -b 32 -r 8k -e filename.raw filenaname.wav play 

If I am starting to get the hang of sox this should just play the input. You as a human being will have no problem recognizing the words. As mentioned above machines do have difficulty with that task.

Now play the input from the mic with these echoes added: echo 11 1 13 1 17 1 19 1 23 1 29 1 31 1 37 1 47 1 53 1.

It will sound like an horrendous mush, but keep listening. After a while you will start to recognize patterns, a while latter you may even be able to convinces your self that you could what is being said. Human beings are very good at learning to recognize patterns. Machines are also able to recognize patterns, when they have to be given examples to match against.

Languages contain many tens of thousands of words, all of them made up of fewer fifty spoken sounds. So use a piece of pattern matching software to match the sounds in sequence, build up the words one sound at a time. I intend to use chatscript for this, but if you are more familiar with aiml or some other language, it should work.

To make the machine understand the words is a bigger problem.

------------------------------------------------------------------------------
Dive into the World of Parallel Programming! The Go Parallel Website,
sponsored by Intel and developed in partnership with Slashdot Media, is your
hub for all things parallel software development, from weekly thought
leadership blogs to news, videos, case studies, tutorials and more. Take a
look and join the conversation now. http://goparallel.sourceforge.net

Re: make spoken words machine recognizable.

[Mike Hamilton]

> a while latter you may even be able to convinces your self that you could
what is being said

It's a while latter, and I'm still not able to convinces myself that I could
what is being said.

> It will sound like an horrendous mush

No argument there, Paul.

> Languages contain many tens of thousands of words, all of them made up of
fewer fifty spoken sounds [...] 

(groan)

> To make the machine understand the words is a bigger problem

Paul, the far bigger problem - vastly beyond our current technology - is to
understand what the blazes you are on about. 

Several people have replied to you in good faith, trying to help, but I for
one now give up. 

We've all said, again and again, that SoX doesn't do speech to text. But as
you've so brilliantly discovered a better way than "very complex
mathematical algorithms to work out what are words and what are not words,
and then to work out what the word might be" by simply applying the Sox echo
effect, you're just too advanced for mere mortals. Patent it, and then sell
it to Nuance (the Dragon Dictate people).

Whatever your astonishing breakthrough may or may not be, it's certainly
nothing to do with SoX. 

Jan Stary, can you spare me some blood pressure tablets - I need them !


------------------------------------------------------------------------------
Dive into the World of Parallel Programming! The Go Parallel Website,
sponsored by Intel and developed in partnership with Slashdot Media, is your
hub for all things parallel software development, from weekly thought
leadership blogs to news, videos, case studies, tutorials and more. Take a
look and join the conversation now. http://goparallel.sourceforge.net

Re: make spoken words machine recognizable.

[Jan Stary]

On Jan 04 16:02:34, 4-werk@gmx.com wrote:
> A better way to explain why I want to add echo to a voice to make spoken words machine recognizable. Current voice to text systems use very complex mathematical algorithms to work out what are words and what are not words, and then to work out what the word might be.

I'll try for the last time: SoX has no such functionality,
whatever magic you think you are doing with the echo effect.

> -d gain -n -D -b 32 -r 8k -e filename.raw filenaname.wav play 
> If I am starting to get the hang of sox this should just play the input.

This is far from even being a valid SoX command line. Apparently,
you haven't bothered to read at least the Introduction in the SoX manual.

> You as a human being will have no problem recognizing the words.
> As mentioned above machines do have difficulty with that task.

Yes. That's called speech recognition. SoX cannot do anything like that.

> Languages contain many tens of thousands of words, all of them made
> up of fewer fifty spoken sounds. So use a piece of pattern matching
> software to match the sounds in sequence, build up the words one sound
> at a time.

Yes. That's called speech recognition. SoX cannot do anything like that.

> To make the machine understand the words is a bigger problem.

Let's start with humans understanding written text: go away.


------------------------------------------------------------------------------
Dive into the World of Parallel Programming! The Go Parallel Website,
sponsored by Intel and developed in partnership with Slashdot Media, is your
hub for all things parallel software development, from weekly thought
leadership blogs to news, videos, case studies, tutorials and more. Take a
look and join the conversation now. http://goparallel.sourceforge.net