Skip to main content

Topic: Comparing codecs by comparing waveforms (Read 4333 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
Comparing codecs by comparing waveforms
I have compared Vorbis to MP3 by:
1. Preparing a WAV file from a CD.
2. Encoding the WAV to MP3 at 192kbps.
3. Encoding the WAV to Vorbis, trying to match the MP3's filesize.
4. Opening the WAV, MP3, and Vorbis in Audacity (audio editor) to check which codec represents the original WAV closest.

So far, Vorbis is closer to the original WAV.
Am I doing this wrong?

  • bandpass
  • [*][*][*][*]
Comparing codecs by comparing waveforms
Reply #1
Yes, very wrong -- these are perceptual audio codecs, and audio is perceived using your ears, not by eyeballing waveform graphs.

  • dhromed
  • [*][*][*][*][*]
Comparing codecs by comparing waveforms
Reply #2
Quote
check which codec represents the original WAV closest


Lossy codecs are designed to take away as much as possible. Matching the waveform picture means nothing at all.

  • db1989
  • [*][*][*][*][*]
  • Global Moderator
Comparing codecs by comparing waveforms
Reply #3
In addition, comparing waveforms in the time domain is often confounding due to perceptually benevolent but visually significant alterations in phase, etc. The frequency domain is more informative. However, neither are considered a valid metric for assessing lossy compression here. Double-blind listening tests are the only recommended method. Please read #8 of the Terms Of Service, as I must assume you have not done so already.

  • Alexxander
  • [*][*][*][*]
Comparing codecs by comparing waveforms
Reply #4
As I read once in some post on this board: You don't listen with your eyes.

  • pdq
  • [*][*][*][*][*]
Comparing codecs by comparing waveforms
Reply #5
Two waveforms can look almost identical but sound very different, of they can look very different but sound almost the same.