[asterisk-speech-rec] New Voice Activity / Barge-in Parameters for use with LumenVox

[Stephen Keller]

As you probably heard if you're a LumenVox customer, we recently
released a new version of our Speech Engine (the 7.5.600 release) that
includes revamped voice activity detection (VAD) technology. VAD is what
lets the Engine separate background noises from a caller's speech. 

The new VAD is significantly better than before. We tested it against a
very noisy set of audio where the old Engine failed to correctly
separate background noise from speech -- the new VAD correctly ignored
the noise 92% of the time. Because background noise and false barge-in
are significant problems for many users, we have seen these changes
improve overall accuracy by several percentage points in some
applications.

As part of these changes, some of the VAD parameters in lumenvox.conf
have changed with. Digium has just released b16 of
res_speech_lumenvox.so (the Asterisk-LumenVox connector module) which
supports these changes.

We have added two new parameters to control barge-in sensitivity:
vad_snr_sensitivity and vad_volume_sensitivity. Both are set on a scale
of 1-100 and default to 50.

vad_snr_sensitivity determines how much louder the speaker must be than
the background noise in order to trigger barge-in. The smaller this
value, the easier it will be to trigger barge-in. 

vad_volume_sensitivity is the volume required to trigger barge-in. The
smaller the value, the more sensitive barge-in will be. This is
primarily used to deal with poor echo cancellation. By setting this
value higher (less sensitive) prompts that are not properly cancelled
will be less likely to falsely cancel barge-in.

We have removed vad_bargein_lvl and vad_noise_floor; if you attempt to
set them using 7.5.600 or later nothing will happen.

If you are a LumenVox user, we strongly recommend you upgrade to the
7.5.600 release of the Speech Engine and b16 of res_speech_lumenvox.
These downloads are available from our Web site at
http://www.lumenvox.com/customers/ by logging into your account and
clicking "Your Software Downloads."

You can find instructions on downloading and installing the software at
http://help.lumenvox.com/?project=lvasterisk

If you haven't upgraded in a while, be aware that res_speech (the
generic Asterisk speech API) changed a few months ago, so if you're
using an older version of Asterisk 1.4 or a precompiled version of 1.2
(e.g. trixbox), you need to make sure you also have the latest version
of res_speech.

At this time, the 7.5 series is not supported on Asterisk Business
Edition. We are working with Digium to get this updated, and I'll be
sure to send a message to the list when this is the case.

I'll be happy to answer any questions about this release.

Stephen Keller
LumenVox Support
P: 877-977-0707, just say "Support"
F: 858-707-7072
Support at LumenVox.com
www.LumenVox.com <http://www.lumenvox.com/> 

 <http://www.lumenvox.com/> 

Winner "Best Innovation in Speech Recognition"
AVIOS SpeechTEK Award