Talkbox, a python scikits for audio signal processing
talkbox is a set of python function for audio signal processing.
The goal is to provide a reasonable set of features often useful for audio
signal processing (speech, music), which are not available yet in scipy.
Another particular feature is that every algorithm will be first implemented
in (as clear as possible python), and in C if needed, so that the toolbox can
be used for educationally purpose. Talkbox in (very) early stage. You can find
the (really rough as for today) documentation here. The targeted features
for a 1.0 release are:
- spectrum analysis: both non-parametric (periodogram, etc...) and parametric (ar, high-resolution, etc...)
- LPC analysis
- Various filters for auditory Modeling, speech feature extraction (mel filterbank, etc...)
- Various transforms like DCT, MDCT, etc...
- MFCC computation
- Formant analysis tools, etc...
I will first implement a sub-set of matlab features available in its signal processing toolbox, and build on that to offer things available in praat, etc...
All the code is available under the very liberal, open source BSD license: the code can be used in both proprietary and open source projects.
Download
For now, only available from svn:
svn co http://svn.scipy.org/svn/scikits/trunk/talkbox
And follow the instructions for installation given in the README.
What's with the picture and the name ?
The talkbox is a famous device to transform its own voice into a robotic kind of voice, and got popular in the 70-ies, used by Stevie Wonder in particular. Roger Troutman made a big hit in the early 80-ies with it, and it has been used heavily in electro-funk since then. Also been used in hip-hop, particularly the G-funk and West-coast hip-hop (Warren-G, Dre, Snoop doggy dog).
It is not the same as a vocoder, although people often mix both devices. I am a big fan of Stevie Wonder and electro-funk, so I thought it would make for a cool name. Maybe sound lame to people unaware of the device, though.