[speex-dev] Higher Bandwidth at lower quality settings

[Trevor Yensen]

Hi Jean-Marc,
        I thought at quality 3 (wideband) - wb_submode1 that the 4-8k band was not
using a codebook table.  From the code I can see that some sort of "lsp"
encoding is performed.  What exactly does this encode?  (I assume lsp means
line-spectral pairs)
        The reason I am asking is I'm comparing the "effective" spectral bandwidth
of Speex against the AMR-WB (G.722.2) codec.  (Available from 3GPP.org)
Speex performs very comparably to AMR-WB until the quality setting drops
below around quality 5.  At approximately equivalent bitrates AMR-WB
maintains the spectral bandwidth better, whereas Speex doesn't seeem to
match above 3.5kHz.  I was wondering if it would be possible to tweak Speex
to drop the narrowband quality at the expense of increasing the bandwidth?

Thanks,
Trevor

<p>-----Original Message-----
From: owner-speex-dev at xiph.org [mailto:owner-speex-dev at xiph.org]On
Behalf Of Jean-Marc Valin
Sent: Thursday, February 27, 2003 12:22 AM
To: speex
Cc: Trevor Yensen
Subject: Re: [speex-dev] Higher Bandwidth at lower quality settings

<p>> 	I was wondering if anyone has experimented with Speex's wideband (16kHz)
> mode at lower quality settings.  In particular I have been using quality
3,
> and with wideband input files the resultant frequency spectrum is limited
to
> about an upper end around 3.5kHz (almost telephony quality bandwidth).
Has
> anyone tried increasing the spectral bandwidth at the expense of lowering
> the narrowband quality (ie using two very low bitrate codebooks, one for
> narrowband and one for the 4-8kHz band - to match the same bitrate at
> quality 3)? If you have can you comment on the quality?  (Looking through
> the Speex code, I noticed that at lower quality settings wideband mode
> basically reverts to narrowband.)

I'm really not sure what you mean. In wideband mode, both the low-band
(0-4 kHz) and the high-band (4-8 kHz) are encoded even at low bit-rates.
When encoding below a certain bit-rate, the high-band becomes a rough
approximation, but it's still there.

        Jean-Marc


--
Jean-Marc Valin, M.Sc.A.
LABORIUS (http://www.gel.usherb.ca/laborius)
Université de Sherbrooke, Québec, Canada

<p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'speex-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.