Control of Voice Activity Detection (VAD)

[aboba]

In WebRTC 1.0 Section 4.2.4 we have:

dictionary RTCOfferOptions {
long offerToReceiveVideo;
long offerToReceiveAudio;
boolean voiceActivityDetection = true;
boolean iceRestart = false;
};

What is the equivalent means by which ORTC API controls voice activity detection?

[aboba]

In WebRTC 1.0 the voiceActivityDetection attribute determines what is put into the Offer (such as inclusion of CN as defined in RFC 3389). In ORTC API, we can control the codecs that are offered (such as CN) and/or the settings of those codecs (e.g. support for comfort noise in G.729). Therefore there is not a need for an explicit voiceActivityDetection attribute within RTCRtpParameters.

[aboba]

While no specification change is required, it would be useful to add a Section 16 on WebRTC compatibility issues, so as to answer this and other potential WebRTC compatibility questions. Here is the proposed text for a Section 16.1 within that section:

16.1 Voice Activity Detection

[WEBRTC10] Section 4.2.4 defines the RTCOfferOptions dictionary, which includes the voiceActivityDetection attribute, which determines whether Voice Activity Detection (VAD) is enabled within the Offer produced by createOffer(). The effect of setting voiceActivityDetection to TRUE is to include the Comfort Noice (CN) codec defined in [RFC3389] within the Offer.

Within ORTC API, equivalent behavior can be obtained by configuring the Comfort Noise codec for use within the RTCRtpParameters object, or configuring a codec with built-in support for Comfort Noise (such as G.729) to enable comfort noise.