Speex dev - Aug 2008 - size of speex packets when VAD/CNG is enabled

Hi all,

supposing to have speex working at 3.95, when VAD/CNG is employed, which 
is the size of those "silence" samples respect to voice samples?

I am wondering if it could be possible to lower even the transmission of 
VAD/CNG samples by specifying "outbound" in the communication channel 
only 1 bit instead of sending the whole VAD/CNG packet.

I am doing the same for AMR where i strip-off the AMR SID describing 
VAD/CNG and simply send "1 bit" in a custom appropriate header of the 
transmission protocol. With AMR 4.75, sample of 13byte, additionally i 
also cut-off the the 1byte AMR header by making AMR samples to be 
transmitted of 12byte.

Are those kind of "bandwidth" optimization feasible also for speex?

a) size of VAD/CNG samples being replaced by "1 bit" instead of 
transporting the whole sample
b) cutting-off the speex header to reduce the sample size (and 
re-attaching it a standard-one on the other side before playing)

I don't know if the first (a) or the second (b) would be feasible or 
would cause problems.

Fabio

p.s. i am talking about a voip application

you can use the marker bit on the RTP.

- farhan

On Tue, Aug 19, 2008 at 12:04 PM, Fabio Pietrosanti (naif)
<lists at infosecurity.ch> wrote:
> Hi all,
>
> supposing to have speex working at 3.95, when VAD/CNG is employed, which
> is the size of those "silence" samples respect to voice samples?
>
> I am wondering if it could be possible to lower even the transmission of
> VAD/CNG samples by specifying "outbound" in the communication
channel
> only 1 bit instead of sending the whole VAD/CNG packet.
>
> I am doing the same for AMR where i strip-off the AMR SID describing
> VAD/CNG and simply send "1 bit" in a custom appropriate header of
the
> transmission protocol. With AMR 4.75, sample of 13byte, additionally i
> also cut-off the the 1byte AMR header by making AMR samples to be
> transmitted of 12byte.
>
> Are those kind of "bandwidth" optimization feasible also for
speex?
>
> a) size of VAD/CNG samples being replaced by "1 bit" instead of
> transporting the whole sample
> b) cutting-off the speex header to reduce the sample size (and
> re-attaching it a standard-one on the other side before playing)
>
> I don't know if the first (a) or the second (b) would be feasible or
> would cause problems.
>
> Fabio
>
> p.s. i am talking about a voip application
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev
>

Yes but it seems to me that in the Speex RTP profile there's no such 
possibility of specifying which kind of sample are inside the RTP packet 
(having multiple sample inside).

If i have 4 sample:
voice sample 1
silence sample 2
voice sample 3
silence sample 4

How the aggregated payload will look like?

I would like to be able to specify somewhere, for example using 4 bit, 
if there's voice or not and putting inside the payload only:
voice sample 1
voice sample 3

I don't know which is the size of the "silence sample" in speex.

In AMR 4.75 it's 6byte for example and it can be reduced to 1 bit.

What about in speex?

Fabio

Ashhar Farhan wrote:
> you can use the marker bit on the RTP.
>
> - farhan
>
> On Tue, Aug 19, 2008 at 12:04 PM, Fabio Pietrosanti (naif)
> <lists at infosecurity.ch> wrote:
>   
>> Hi all,
>>
>> supposing to have speex working at 3.95, when VAD/CNG is employed,
which
>> is the size of those "silence" samples respect to voice
samples?
>>
>> I am wondering if it could be possible to lower even the transmission
of
>> VAD/CNG samples by specifying "outbound" in the communication
channel
>> only 1 bit instead of sending the whole VAD/CNG packet.
>>
>> I am doing the same for AMR where i strip-off the AMR SID describing
>> VAD/CNG and simply send "1 bit" in a custom appropriate
header of the
>> transmission protocol. With AMR 4.75, sample of 13byte, additionally i
>> also cut-off the the 1byte AMR header by making AMR samples to be
>> transmitted of 12byte.
>>
>> Are those kind of "bandwidth" optimization feasible also for
speex?
>>
>> a) size of VAD/CNG samples being replaced by "1 bit" instead
of
>> transporting the whole sample
>> b) cutting-off the speex header to reduce the sample size (and
>> re-attaching it a standard-one on the other side before playing)
>>
>> I don't know if the first (a) or the second (b) would be feasible
or
>> would cause problems.
>>
>> Fabio
>>
>> p.s. i am talking about a voip application
>> _______________________________________________
>> Speex-dev mailing list
>> Speex-dev at xiph.org
>> http://lists.xiph.org/mailman/listinfo/speex-dev
>>
>>