Journal Spy der Mann's Journal: Revamping MP3: An idea that might just work 1
I was discussing with a friend about how 160 or 192kbps mp3 can be heard much clearer (they sound more soothing to the ear) than 128kbps mp3's.
Then I told him about Coldplay Clocks and how the notes got distorted because of the missing harmonics. And I thought:
Some waveforms usually have harmonics at regular intervals. So, why not store an algorithm for these intervals? Like, instead of saying, "there's a peak at A,B,C,D and E", say "there are peaks starting at A every N khz, decreasing by M db's"?
I think neural network algorithms can be implemented. I've done General Regression Neural Networks (they're easy) to do interpolation.
How effective would this measure be? I don't know. Perhaps there are already more efficient algorithms to compress audio, but they've been ignored due to MP3's popularity.
Sometimes de-facto standards have just too much weight.
Ogg? (Score:2)
Ogg Vorbis [vorbis.com] seems like a contender. I don't have the ear to distinguish the difference, but those who do tend to say that it is able to provide a higher quality than MP3. If you haven't compared, you might look into it.
MP3 being the de facto standard is a problem that you've hit squarely on the head. It certainly isn't going away any time soon, and since there is money to be made by selling it, it gets pushed a lot harder than ogg. The good news is that the main Linux distributions are pushing for it since