Gstreamer elements implementing some suport for rendering Audio Description.

The elements are,

• whp198dec - extracts AD_descriptor structures from an audio waveform, encoded per BBC R&D whitepaper WHP 198
• adcontrol - consumes buffers of AD_descriptor structures and uses these to control an internal 'volume' element; used to implement the 'fading' of the audio of the main presentation as required for the audio description content to be heard clearly

                   +-------------+
    WHP 198 signal |             |
    +-------------->  whp198dec  |
                   |             |
                   +-------------+
                          |
                          | AD_descriptors
                          |
                   +------v------+
    Main audio     |             |  Main audio (with fade)
    +-------------->  adcontrol  +--------------->
                   |             |
                   +-------------+
                                    Description audio
    +-------------------------------------------->

Limitations

• Volume changes are not explicitly queued to the match audio stream, which might cause problems for some pipeline structures (untested)
• The whp198dec element hs not been generalized to support multiple sample formats and bit rates - use other Gstreamer elements to convert as required
• Ignores 'pan' information (I have no example content using the panning feature)

Example pipeline

Given a test.wav contains description in the left stereo channel, and the WHP 198 control data in the right channel, this pipeline plays the description track using a noise test signal for the main audio, as a basic demo of the control over the main audio's volume level.

gst-launch-1.0 \
  filesrc location=test.wav \
	! wavparse \
	! deinterleave name=d \
  d.src_1 \
	! queue max-size-time=100000000 \
	! audioconvert \
	! audio/x-raw,format=S16LE,rate=48000,channels=1 \
	! whp198dec \
	! ad. \
  d.src_0 \
	! queue max-size-time=100000000 \
	! audioconvert \
	! audio/x-raw,format=S16LE,rate=48000,channels=1 \
	! mix. \
  audiotestsrc wave=red-noise volume=0.3 \
	! audio/x-raw,format=S16LE,rate=48000,channels=1 \
	! adcontrol name=ad \
	! mix. \
  audiomixer name=mix \
	! autoaudiosink