Acceptable speech to noise ratios

Hello,

My husband works in the audio industry so I asked him to comment on the "20 
dB" figure cited in Guideline 1.7 [1], level 2 success criterion 1:
""
Audio content does not contain background sounds OR the background
sounds are at least 20 decibels lower than the foreground audio content.

Note:

A 20 db difference in sound level is roughly 4 times quieter (or louder).
""

His email response follows. He references a resource [2] from Shure 
Brothers that describes how to record speech so that it will be 
intelligible. Note that the success criterion implies a 20dB difference 
when the user hears the audio and the Shure reference describes recording 
of speech (rather than the difference that the user might hear when 
listening to the playback).  It says, "If speech level is 20dB to 30dB 
above background noise level, intelligibility will be fair to good.  If 
speech level is 30dB to 50dB above background noise level, intelligibility 
will be good to excellent."  Perhaps we ought to use 30dB - 50dB instead of 
20dB?  Is there a resource that demonstrates that 20dB is sufficient?

There is also a white paper, "Understanding Sound System Design and 
Feedback Using (Ugh!) Math" that has more detail. [3]

[1] http://www.w3.org/WAI/GL/WCAG20/#audio-contrast
[2] http://www.shure.com/support/technotes/app-snratio.html
[3] http://shure.com/pdf/booklets/sound_system_design.pdf

>From: "John Boudreau"
>To: "Wendy A Chisholm" <wendy@w3.org>
>
>The topic is typically called direct sound to Ambient noise ratio:
>Here's a quick guide to what's acceptable for speech in a room from our 
>friends at shure
>http://www.shure.com/support/technotes/app-snratio.html
>20dB is the minimum acceptable Signal to Noise ratio.  40 to 50dB is 
>optimal and should be strived for.
>
>As far as 20db being 4x quieter.  This is mostly true.  The "rule of 
>thumb" for dB references go as follows:
>1dB - the unit of perceivable change in sound level to the human ear
>3dB - generally considered the realistic change in sound level that humans 
>can hear in everyday environments
>6dB - significant change in level.  6dB is also significant because if you 
>measure SPL at x distance and then measure it again at 2x distance it will 
>drop 6dBSPL (inverse square law)
>10dB - twice as loud or half as loud
>you could logically say 20dB will be 4x quieter.
>
>These are simply reference points and markers for ez recall by us 
>humans.  That's why I say rule of thumb, as they are generally true but 
>extremely averaged number references.  Any individual in a given acoustic 
>environment could make measurements that might not agree.

-- 
wendy a chisholm
world wide web consortium
web accessibility initiative
http://www.w3.org/WAI/
/--

Received on Thursday, 6 November 2003 18:53:21 UTC