Diva Tech Meeting
2009.Jun.12
Minutes taken by Karl

Cam:

• Cam has put the project into the SVN repository in a directory called DIVA.
• Cam has almost pulled together the different pieces to go all the way from the glove to sound. At the moment the output is going to a .wav file. He is in the process of figuring out the RtAudio object from Perry's STK. He is hoping to get sound out of it today (cross fingers).

Johnty:

• Johnty has finished writing the drivers and they are ready for Cam.
• There are some adjustments that happen inside the polhemus drivers to level the orientation of the axes. There was some debate as to whether those adjustments (and other possible adjustments) should be happening inside the driver or outside. The original rationale was to have it inside to make the code execute faster in C++. Now that everything is C++, this can be reconsidered. Johnty is further looking into it to see what makes most sense.

Allison:

• Allison has been busy finishing up a project for a course from the previous term.
• She asked whether we should be using holmes_c++ for Liz and I confirmed that, yes, this is what we want. I believe that the holmes_c++ version is most stable. It would be good to verify this by using it over time rather than waiting for a live performance.

Naming issue:

• the new Diva is checked into a directory called DIVA, which seemed fine at first. We now have a directory structure that looks like this: /DIVA/src/DIVA/. This could get confusing.
• we tossed around different ideas for renaming the system. Diva 2, Diva 3. Naming after birds.

Karl:

• Holmes designed the formant synthesizer to have 12 parameters, but we are only using 11 of them. The additional parameter controls how long the vocal folds are open (commonly referred to as open quotient). Holmes refers to this as the Mark-Space (MS) parameter. I adjusted the MS parameter and the difference is dramatic. You may not hear any difference through laptop speakers, but the results are dramatic through headphones. I have attached sample files to this email.
• In the vowels we are using, there are no long or short A sounds (as perceived by my Canadian ears). They all sound like E's. I'm going to try to find better formant values for the vowels in the system. I know that other people people have given a kick at this, and now it's my turn to give it a shot. I'm in the process of figuring out how to test my results.
• I have been doing a lot of code refactoring to make the meaning of the code far more clear.