See http://www.ietf.org/mail-archive/web/speechsc/current/msg01264.html
hi, 
when server receives a queued RECOGNIZE, it sends a response to client with 
PENDING request-state. When actual RECOGNIZE ends, server signal event to client 
with RECOGNITION-COMPLETE event, but has actually no event to signal to the 
client that new RECOGNIZE has started. 

In Synthesis resource, this work is done by SPEECH-MARKER event with marker 
value="", in Recognition resource, this event is missing.

Similar question is present on Verification resource, when both recognition and 
verification resources shares the same session: when server receives a VERIFY, 
it send a PENDING response to client (it's waiting for the next RECOGNIZE), but 
when it receives RECOGNIZE, it has no event to signal to client that VERIFY has 
moved to IN-PROGRESS-STATE.

In order to solve this problem, we should define some event as RECOGNITION-
STARTED and VERIFICATION-STARTED, which have the meaning of the moment in which 
audio samples can be sent to the recognition resource.