Frequently Asked Questions (FAQs):Quality of Calls - Voice Activity Detection (VAD)

VoIP: Quality of Calls - Voice Activity Detection (VAD)

To Reduce the amount of Data sent throught the network, most IP-systems dispose of an optional mechanise to detect whether the current frame contains speech. Only if it does , is it encoded and sent across the network. In case no speech is sent, a comfort noise is commonly inserted in the periods of silence at the receive side so that the channel is not entirely muted, and the user is assured that the call is still active.

Voice activity detection (VAD) and comfort-noise generation may both lead to a perceivable speech quality degredation.

For example , VAD may cause clipping and hence loss of information.
In addition, VAD interacts with the effect of packet loss:
On one hand, with VAD the loss distribution and rate will be different because less data is sent across the network.
On the other hand, packet loss in periods of no speech in the connection without VAD may lead to artefacts on the background noise transmitted from the send side, which could be avoided with VAD.

We suggest that if you are experiencing clipping that you turn off VAD on your VoIP device/software.