Newsgroups: comp.dsp,comp.speech
From: joop@entropic.demon.co.uk (Joop Jansen)
Path: lyra.csx.cam.ac.uk!pipex!demon!entropic.demon.co.uk!joop
Subject: Re: Bark spectrum, Bark scale
References: <chr.1.000E4CAE@cec.comm.se>
Organization: Entropic Research Laboratories, Inc
Reply-To: joop@wrl.epi.com
X-Newsreader: Demon Internet Simple News v1.29
Lines: 31
Date: Thu, 25 Aug 1994 17:57:58 +0000
Message-ID: <777837478snz@entropic.demon.co.uk>
Sender: usenet@demon.co.uk
Xref: lyra.csx.cam.ac.uk comp.dsp:9095 comp.speech:3142

In article <chr.1.000E4CAE@cec.comm.se> chr@cec.comm.se "Christian Ronne" writes:

> I am dealing with objective quality evaluation of speech coders. One method 
> uses BARK SPECTRUM and also mentions BARK SCALE.
> 
> I am not quite sure what these terms mean. Can anyone help me?
> 

Hello Christian,

The Bark scale is a logarithmic-like mapping of the frequency axis to a 
new scale. The mapping takes into account the way humans perceive sound. 
The human ear is less sensitive to frequency changes in the higher frequency
range than in the lower frequency range when you are on a Herz scale. When
you are on a Bark scale equal steps will mean equal difference in
perception.

An article of Hynek Hermanski in JASA "Perceptual linear prediction analysis
of speech" 87 (4), April 1990 describes the implementation among other
things.

Best regards Joop Jansen

-- 
Entropic Research Laboratories, Inc
Sheraton House
Castle Park
Cambridge CB3 0AX, UK
E-mail: joop@entropic.com
Tel: +44-223-302651
Fax: +44-223-324560
