5.3.1 Fine-Tuning VAD Performance on SpringWare Boards

On SpringWare boards, several of the defines used for VAD can be divided into two groups. You can adjust the values of these two sets of defines to fine-tune the performance of the VAD.

• Defines that take effect during the playing of a prompt:
• DXCH_SPEECHPLAYTHRESH
• DXCH_SPEECHPLAYTRIGG
• DXCH_SPEECHPLAYWINDOW
• Defines that take effect after play has completed:
• DXCH_SPEECHNONPLAYTHRESH
• DXCH_SPEECHNONPLAYTRIGG
• DXCH_SPEECHNONPLAYWINDOW

During play, echo or noise often exists on the channel. If you find that your application triggers on echo or background noise, you may want to:

• increase speech trigger (DXCH_SPEECHPLAYTRIGG) so that the incoming speech energy is present for a greater duration of the speech window.
• increase speech window (DXCH_SPEECHPLAYWINDOW) thereby requiring the incoming speech energy to be present for a longer time period. This should protect against false triggers due to noise spikes or other short duration (non-speech) noises.
• increase speech threshold (DXCH_SPEECHPLAYTHRESH) so that the energy level of incoming speech required to trigger the VAD is relatively higher during play and relatively lower when the prompt play completes (DXCH_SPEECHNONPLAYTHRESH).

Note:

To reduce sensitivity to background noise, you must perform these actions in certain combinations only: increase speech trigger alone; increase speech threshold alone; increase speech trigger and speech window together; increase speech trigger, speech window, and speech threshold together. Increasing speech window alone will not help reduce sensitivity to background noise.

After the prompt completes, residual echo and VAD sensitivity to prompt-related false triggers should be reduced. Hence, you may set speech trigger (DXCH_SPEECHNONPLAYTRIGG), speech window (DXCH_SPEECHNONPLAYWINDOW) and speech threshold (DXCH_SPEECHNONPLAYTHRESH) to lower values allowing for easier speech detection. Doing so improves rejection of false triggers and VAD sensitivity.

For more information on parameters, see the ec_setparm() function description in the Continuous Speech Processing API Library Reference.

Click here to contact Dialogic Customer Engineering