A compression algorithm concept

CreatorLes · Joined: Oct 05, 2014 Posts: 84 Location: San Antonio TX USA

Hi, to win a contest I am trying to develop a compression algorithm. I was thinking that since phase is not detected by the human ear, we could sort of throw phase out the window and just save the amplitude spectrum of the FFT over time as the compression technique, then compress even THAT reduced thingie with a conventional compression technique.

Is this already done? Would it work?

Les

DrJustice · Posted: Sun Nov 16, 2014 8:22 am Post subject:

You need to preserve the phases since they are what give us the precise frequencies, i.e. when frequencies in the signal don't land on the exact bin frequencies. This can be seen as a phase shift that changes between the successive spectral frames.

If you used an extremely high number of bins (10k or more), that might help some, but I fear that it still wouldn't sound good. The overlaps would also be huge in order to maintain good temporal resolution. Thus it would be bad for the compression side of things. The phases are needed both for fidelity and for compactness of the data.

CreatorLes · Joined: Oct 05, 2014 Posts: 84 Location: San Antonio TX USA

OK, scratch that idea, I'll tell you what I am going for here. I want to win a 3D printing contest sponsored by Ghost Vinyl Records on www.thingiverse.com and I was thinking if i made a 45 record that would be cool. Then I thought, it would cost a lot to get a USB record player and accessories so why not make some custom thing?

I have high quality cams on my MacBook and iPhone, so how about if I 3D print some kind of encoded panel that records a song. You'd hold it up and snappy a photo of it to convert to mp3. I could write the mp3 to 3d print and image of 3d print to mp3 conversion software in Processing, which I have become somewhat proficient at programming.

So any thoughts or ideas along this line are welcome!

Les