Prev: Getting an houlry average from a data set
Next: Compiling Toolbox Material
From: Turner on 22 Jun 2010 00:18
I'm writing a simple program that I would like to control with verbal commands. I've read what I can understand about voice command. I'm finishing my M.S. Mechanical Engineering. This will be for a single user, 20 word vocabulary. The user will be trained to use a specific syntax to help with recognition. For example: User says, "Computer" to initialize the system to listen for the following commands. Then user says,"range 7 2 0", then matlab would compare to the dictionary and reply, "range 7 2 0". I have used the txt2speach.m so a simple output to the screen would be fine.

Could someone write a simple m-file that allows people to speak numbers 0-9 to matlab and matlab out puts them on the screen? I think it would be a cool demo that would help start more voice projects.

This could be a matlab contest where the winner is determined by the fastest spoken echo response to a vocabulary of 10 words and numbers 0-9. My goal would be code that is 25-50% as good as the UPS tracking number call center. I can speak the tracking number faster than a person could write and UPS always gets it right!
From: Steven Lord on 22 Jun 2010 11:40

"Turner " <turner.j.harris(a)gmail.com> wrote in message
news:hvpdhs$g78$1(a)fred.mathworks.com...
> I'm writing a simple program that I would like to control with verbal
> commands. I've read what I can understand about voice command. I'm
> finishing my M.S. Mechanical Engineering. This will be for a single user,
> 20 word vocabulary. The user will be trained to use a specific syntax to
> help with recognition. For example: User says, "Computer" to initialize
> the system to listen for the following commands. Then user says,"range 7 2
> 0", then matlab would compare to the dictionary and reply, "range 7 2 0".
> I have used the txt2speach.m so a simple output to the screen would be
> fine.
>
> Could someone write a simple m-file that allows people to speak numbers
> 0-9 to matlab and matlab out puts them on the screen? I think it would be
> a cool demo that would help start more voice projects.

Sure, once you're done please post it to the File Exchange. But keep in
mind I don't think it's going to be a "simple m-file" -- even humans, who
have had a LOT of time to practice this skill, sometimes don't get it right
(particularly where there are accents [is "cah" actually "car" with a Boston
accent or a crow's call with a slightly different accent?] or homophones [to
versus too versus two] involved.)

http://en.wikipedia.org/wiki/Boston_English

http://en.wikipedia.org/wiki/Homophone

> This could be a matlab contest where the winner is determined by the
> fastest spoken echo response to a vocabulary of 10 words and numbers 0-9.
> My goal would be code that is 25-50% as good as the UPS tracking number
> call center. I can speak the tracking number faster than a person could
> write and UPS always gets it right!

I'm not so sure it would be a good MATLAB contest -- the problem is
vulnerable to test suite deduction a la Mastermind (my first entry returns 0
0 0 0 0, my second 1 1 1 1 1, my third 2 2 2 2 2, etc. until I know what
words are in the phrase, then I start permuting.) In addition, I believe
that the UPS tracking number call center has the advantage of a major
assumption: all the entries you speak to it are numbers.

--
Steve Lord
slord(a)mathworks.com
comp.soft-sys.matlab (CSSM) FAQ: http://matlabwiki.mathworks.com/MATLAB_FAQ
To contact Technical Support use the Contact Us link on
http://www.mathworks.com


 | 
Pages: 1
Prev: Getting an houlry average from a data set
Next: Compiling Toolbox Material