Sorry, to let me make this more clear. To use a white training sequnece
the absence of DT.
I understand that this echocancel group consist of people who are
involved in compiling the ITU-G.168 standard. My question is addressed to
these members in particular.
I am working on network echo cancellers. Hybrid echo paths have a typical
active region of 5ms-7ms (G.168) and when a call is made, the hybrid echo
path is formed and remains pretty much the same throughout the call.
My question is this;
Why can't we use a training sequence across the hybrid echo path at the
beginning of every call?. Training sequences are used in mobile
communication systems to identify channel characteristics successfully. We
could also use this idea in network echo cancellation to measure the hybrid
echo path. From this measurement, we could identify the active region of
the echo path then, we could use a 5ms-7ms adaptive echo canceller instead
of the usual 64ms adaptive echo canceller (i.e., reducing the complexity of
the echo canceller). Furthermore, the echo cancellers can be made to become
very robust to double talk situations.
Therefore, why don't we include an agreed training sequence into all the
network echo cancellers?. Please let me know all your opinions on this