Minor thirds and background noise (Brian Gygi )

Subject: Minor thirds and background noise
From:    Brian Gygi  <bgygi(at)EBIRE.ORG>
Date:    Sun, 30 Jan 2005 13:38:53 -0800

In general, environmental sounds have a 1/f spectrum, which was the subject
of much discussion on this list a couple of years ago.  However, the
spectral shape of an individual instance of masking background can vary
widely, from speech-shaped noise to broadband noise or, theoretically, even
harmonic (say, if you were sitting in the middle of a horn section).

At 12:46 AM 1/29/2005, Christian Kaernbach wrote:
>Dear List,
>
>In a very speculative way one could assume that one reason for
>minor-third calling is to avoid possible masking by unexpected
>environmental sounds. Assuming that many environmental sounds have a
>spectrum varying strongly as a function of frequency (that is with the
>amplitude in one frequency band not correlated strongly to the amplitude
>in other frequency bands) it would be advisable to code your message
>such that it involves two different frequency bands so that if one of
>these two bands is masked only half of the message is lost. That would
>be better than a 50% chance that the entire message is lost because
>speech is redundant, and because the recipient might solicit a resend of
>the message.