1.3.1 How DirectPlay Handles Voice Bursts
A voice burst occurs when DirectPlay detects voice. Essentially, a single voice burst translates to a "segment of voice recorded sequentially". Multiple voice bursts simply mean that after a burst there is silence. When voice data is transmitted again, that transmittal is the next voice burst.
For example:
Begin recording
Two seconds of audio data
End recording
One-second pause
Begin recording
Three second of audio data
End recording
The preceding scenario generates two voice bursts. The first voice burst is 2 seconds of audio time and the second is 3 seconds of audio. The timing within an audio segment is preserved. The 1 second between bursts (per the preceding example) is not preserved. As a result, the time between voice bursts can end up being 0.8 seconds, or 1.5 seconds, or some other length, depending on network conditions.
All audio data within a voice burst is continuous. Each voice burst is equivalent to a single, continuous buffer of audio. When no audible audio (silence) is recorded, no transmission is necessary. Silence represents the time "between" voice bursts. DirectPlay continues to record/send audio 400 milliseconds (ms) after a user stops speaking. This allows for up to 400 ms of "silence" audio to be sent between words/sentences that the user speaks. This is not required in any way, but buffering in this manner allows for a more continuous transmission of voice data. The algorithm used to begin/end the recording of audio data can be determined by any implementer writing to the DirectPlay Voice Protocol.
The DirectPlay Voice Protocol analyzes audio data to determine whether it needs to be sent. Essentially, this is accomplished by checking to see if there is any audio data over a certain threshold. Determining when a voice burst needs to start is left to the implementation. This is commonly accomplished by having the user press a button on the computer, such as "Press to Talk". The exact manner in which the start of the voice burst is determined is of no consequence to the DirectPlay Voice Protocol or the implementation. As long as an audio burst is a continuous stream of audio data, the data is considered to be in a voice burst. If two audio samples are appended when they are not sequential, this is not a true voice burst because the audio is not continuous.
A voice burst can be arbitrarily long, even indefinite, as determined by the implementation. However, the longer the voice burst, the more likely that audio data will be lost, resulting in "pops" or "skips" in the playback.