Audacity Forum

Update: I've modified the format slightly, and now it not only sounds better, but has better compression (26.67% [frame size = 15] or 28.57% [frame size = 7]). What I ended up doing was scrapping the 'multiple prediction models' idea and using a single model for all the data, and increase the order ...

Okay, I think I've got it. What I ended up doing was: -Use recursion instead of an inverted matrix to solve a system of linear equations [O(n^2) vs O(n^3)] -Calculate linear prediction coefficients for every frame [frame size is now 256 samples] as well as correlation data [this is important] -To me...

Ah. Well, in most LPC codecs, the signal is predicted with no error correction (because it's normally implemented for speech, and you rarely need error correction for it to be recognizable). I account for this in my code (using 4bit nibbles as the correction data and a 2^n scalar for each sample fra...

Ah, I see. Thanks for the links - for some reason, it didn't cross my mind to go to DSP-math groups >_< I'll give it a shot and, if successful, I'll post the results here in case there are others like me looking for a solution to this. On a semi-related note: I wonder if it'd be at all useful to sta...

Okay, I've found a mathematically correct, better way. Instead of quantizing the data using the previous methods, I'll do the following: //! Calculate dot product of two vectors double DotProduct(double *v0, double *v1) { double dot = 0.0; for(int i=0;i<ORDER;i++) dot += v0[i] * v1[i]; return dot; }...

Well, I've been reading some more and have come up with three solutions: 1) Subset/feature selection Apparently, this is the more 'mathematically correct' way of doing it. It's also very computationally expensive and WAY past me. 2) k-d trees In image palette quantization, you can use an octree to r...

Oops, sorry that I didn't see the reply - I didn't receive an e-mail notification for some reason [will check into this in a bit]. Yeah, I've read the Wiki articles, read at least another six articles found on Google, but I still can't figure it out. I think the main issue with my inability to fully...

Hey guys. Newbie here. Sorry that my first post is a request for help but I'm almost literally going insane over this. I've been programming a sound format that achieves 32:9 compression by the use of linear prediction. It's basically a challenge to myself to see whether a format that achieves such ...