Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!agate!howland.reston.ans.net!newsserver.jvnc.net!newsserver.egr.uri.edu!ramli
From: ramli@ele.uri.edu (Ramli)
Subject: Artifacts introduced by filtering
Message-ID: <C9I5Kv.43H@egr.uri.edu>
Summary: Artifacts introduced by filter-bank method
Keywords: filter-bank, processing artifacts
Sender: news@egr.uri.edu (USENET News System)
Organization: URI Department of Electrical Engineering
Date: Thu, 1 Jul 1993 20:14:54 GMT
Lines: 32

I was reading chapter 37 in "Visual Representations of
Speech Signals" edited by Cooke, Beet, and Crawford.  It
is written by Tsopanoglou, Mourjopoulos, and Kokkinakis
and entitled, "Speech representation and analysis by the use
of instantaneous frequency."

In the Introduction section, second paragraph, the authors
state that tracking instantaneous frequency (IF) is a very useful
way to characterize speech signal.  They then point out that
IF makes sense only for narrowband signals while speech is
essentially broadband.  "This problem can be avoided by introducing
appropriate division in the frequency domain using a filterbank.
However, as is well known, such a choice may lead to other
processing artifacts and compromises, and also increase
significantly processing time and system complexity."

What are the processing artifacts that are introduced if
one employs a filter-bank to analyze speech ?  If the filters
are linear phase (FIR), I would think there would be practically
no distortion introduced in the components that are in the
passband of the filter.

I would greatly appreciate if someone can throw more light
on the kinds of artifacts introduced by filter-bank processing
and also give any references that will prove useful.

Thanks!

Ramli.



