Subject: Re: soundscape analysis? Get real!
From: Linda A. Seltzer (lseltzer@phoenix.Princeton.EDU)
Date: Wed Dec 15 1999 - 17:43:27 EST
There are differences in what can be learned from
spectrograms as compared to listening. Spectrograms
have been used for speech analysis for a long time,
and the techniques for music are similar.
1. Spectrograms were developed to look at the details
of particular phonemes. Before we could look at them
on the screens of workstations, the old spectrograph
machines made a narrowband spectrogram for a speech
segment 2.5 seconds long and a wider band spectrogram
for segments twice that in duration. Within that time
period, the details of a few phonemes are examined.
Normally the listener doesn't intend to study by ear
analyzing such short segment, unless one is transcribing
music or taking a diagnostic rhyme test.
2. The listener probably isn't able to consciously
determine what harmonics are present and which are
strongest, but a spectrogram can do this for a few
3. The spectrogram will be more reliable in informing
one about subharmonics which temporarily occur, or
places where the fundamental disappears.
4. Just because some characteristic appears in a
spectrogram of a sound, with all of the other
complexities of a natural sound, you can't assume that
you can go back and do the same thing in digital
synthesis of sounds, without all of those complexities,
and expect it to produce the same effect audibly.
This archive was generated by hypermail 2b27 : Wed Jun 11 2003 - 13:09:17 EDT