Jump to content
  • Advertisement
Sign in to follow this  
ArchangelMorph

speech recognition

This topic is 4015 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I'm hoping to build a tool which I can use to analyse sound files off-line and produce a set of visemes in an output file to be used for simple lip syncing.. The character we have doesn't use blend shapes for facial animation and so the only syncing needed would be having the mouth flap open or closed by a certain amount (between 0 & 1) depending on the speech.. I just have a few questions:- - Should I use a one pole low pass filter and envelope follower & would this give me precise enough results to do what I want to do? - Should I be leaning towards some kind of spectral frequency analysis/ FFT & how long would developing this take generally? I think I have about three weeks max on this so any kind of ball park time estimates would be helpful.. Also if anyone knows any useful resources I can look up to get some info on the two techniques then that would be h'awesome!! Thanks GameDev! :D

Share this post


Link to post
Share on other sites
Advertisement
It's all a matter of how nice you want it to look... The simplest approach is the mouth flapping with the amplitude. Typically you'd have a mouth shape per phoneme and blend them to make it look reasonable. I'm sure there are premade softwares you can use for this sort of thing as unless you want a trivial implementation... I suspect it'll be quite a complex job. I'm certain some sort of spectral method would be necessary to get it working at all (though there are plenty of premade FFT/Wavelet solutions out there) but it all comes down to pattern recognition in a very noise signal. In short, it's not as easy as it might sound.
Hope this helps,

Dan

Share this post


Link to post
Share on other sites
I'm also developing now in direction of speech recognition. For spectral analysis i use the library from : fftw.org
It's easy to use, especially when you only need the FFT function.
But more important is the pattern recognition as mrcheesewheel wrote.. It's not that easy.. I'm myself on the way to find a way :D

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!