Reply by Vladimir Vassilevsky●October 18, 20072007-10-18
Jerry Avins wrote:
>> Hi, because that ADPCM codec is standardized. We can only try to play
>> some tricks to cheat the user's ear, not to convince the standard body
>> to chnage their codec.
>
>
> It seems that you believe there is only one ADPCM standard. There are
> many. If you control both ends, you can choose an appropriate one.
>
> The ITU defines several speech codecs which use ADPCM:
>
> Codec Sampling
> G. Rate
> Number (kHz) Bit Rates (kBps)
> G.721 8 32
> G.726 8 16, 24, 32, 40, 64
> G.727 8 16, 24, 32, 40, 64
There is also G.722 at 16kHz/64kbps as well as IMA and MS ADPCM scalable
to whatever sample rate (4 bits per sample). But any kind of ADPCM
requires at least 32kbps to produce the toll quality speech. Below
32kbps, the ADPCM sound is nasty. It seems like the OP got stuck with
some particular codec realization.
Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Reply by Jerry Avins●October 18, 20072007-10-18
heliboy2003@yahoo.com.tw wrote:
>> Also, I think Vladimir's question is excellent: "Why reinvent the
>> wheels?"
>> --
>
> Hi, because that ADPCM codec is standardized. We can only try to play
> some tricks to cheat the user's ear, not to convince the standard body
> to chnage their codec.
It seems that you believe there is only one ADPCM standard. There are
many. If you control both ends, you can choose an appropriate one.
The ITU defines several speech codecs which use ADPCM:
Codec Sampling
G. Rate
Number (kHz) Bit Rates (kBps)
G.721 8 32
G.726 8 16, 24, 32, 40, 64
G.727 8 16, 24, 32, 40, 64
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by Randy Yates●October 18, 20072007-10-18
heliboy2003@yahoo.com.tw writes:
>> Also, I think Vladimir's question is excellent: "Why reinvent the
>> wheels?"
>> --
>
> Hi, the ADPCM codec is standardized, we have no freedom to either
> modify it or use other better codecs.
> Unless I can convince my boss to send a proposal to persuade standard
> body to change to a better codec.
What is your application?
--
% Randy Yates % "Though you ride on the wheels of tomorrow,
%% Fuquay-Varina, NC % you still wander the fields of your
%%% 919-577-9882 % sorrow."
%%%% <yates@ieee.org> % '21st Century Man', *Time*, ELO
http://www.digitalsignallabs.com
Reply by ●October 18, 20072007-10-18
>
> Also, I think Vladimir's question is excellent: "Why reinvent the
> wheels?"
> --
Hi, because that ADPCM codec is standardized. We can only try to play
some tricks to cheat the user's ear, not to convince the standard body
to chnage their codec.
Reply by ●October 18, 20072007-10-18
> Also, I think Vladimir's question is excellent: "Why reinvent the
> wheels?"
> --
Hi, the ADPCM codec is standardized, we have no freedom to either
modify it or use other better codecs.
Unless I can convince my boss to send a proposal to persuade standard
body to change to a better codec.
Reply by Randy Yates●October 17, 20072007-10-17
heliboy2003@yahoo.com.tw writes:
>> I assumed the definition of "improve the speech quality" meant "improve
>> the listening experience of a person".
>
>
> Exactly!! I am thinking that adding some audio effects (such like
> flanger) to the decoded speech could masked out some of the embedded
> noise resulting from the encoding/decoding (ADPCM) process, so that
> people can feel the decoded speech better than without adding such
> effects. I just don't know if adding such audio effect is a common
> practice or just I am too stupid and bold to think that.
I think that's a horrible idea and a horrible philosophy. Why not
apply the extra processing to improving the fundamental codec? For
example, with a higher sampling rate you could reduce the noise.
Also, I think Vladimir's question is excellent: "Why reinvent the
wheels?"
--
% Randy Yates % "She has an IQ of 1001, she has a jumpsuit
%% Fuquay-Varina, NC % on, and she's also a telephone."
%%% 919-577-9882 %
%%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO
http://www.digitalsignallabs.com
Reply by Vladimir Vassilevsky●October 17, 20072007-10-17
<heliboy2003@yahoo.com.tw> wrote in message
news:1192586067.094892.27230@z24g2000prh.googlegroups.com...
> Hi,
>
> I am in a process to improve the speech quality out of our current
> speech codec. Basically, our speech codec is an ADPCM codec
[...]
It is well known that ADPCM sucks, especially for the low bit rate. Why
reinventing the wheels and not using any of the standard speech coders?
Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com
Reply by ●October 17, 20072007-10-17
> I assumed the definition of "improve the speech quality" meant "improve
> the listening experience of a person".
Exactly!! I am thinking that adding some audio effects (such like
flanger) to the decoded speech could masked out some of the embedded
noise resulting from the encoding/decoding (ADPCM) process, so that
people can feel the decoded speech better than without adding such
effects. I just don't know if adding such audio effect is a common
practice or just I am too stupid and bold to think that.
Reply by Richard Owlett●October 17, 20072007-10-17
heliboy2003@yahoo.com.tw wrote:
> Hi,
>
> I am in a process to improve the speech quality out of our current
> speech codec. Basically, our speech codec is an ADPCM codec with bunch
> of interpolation and decimation filters and Sigma-Delta ADDA, and we
> have already adjusted all the filters so now it works just fine.
> However, we are still not 100% satisfied with the audio quality, we
> want to make the sound more clear so we are thinking of using some
> audio processing methods (beyond filter coefficients adjustment but
> not going to too complex method such like noise reduction ) to achieve
> that. I am thinking of using such like Equalizer, Dithering, Echo,
> Flager.... to try to make the speech better. So my question is what
> are the common ways to achieve that or I am just expecting too much?
> Thanks
>
I noticed that you asked the same question in comp.speech.research.
In neither did you explicitly state that that the codec was to be used
in a speech recognition system. When reading the comp.dsp post, I
assumed the definition of "improve the speech quality" meant "improve
the listening experience of a person".