webrtc/tools/e2e_quality/audio/README
andrew@webrtc.org f4c6aa2e81 Improve the reliablity of the audio e2e test.
- Use higher quality resampling.
- Add a longer delay before pacat recording.

TBR=kjellander@webrtc.org
BUG=issue502
TEST=manually

Review URL: https://webrtc-codereview.appspot.com/646005

git-svn-id: http://webrtc.googlecode.com/svn/trunk@2374 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-07 04:00:15 +00:00

28 lines
1.2 KiB
Plaintext

The tools here run an end-to-end audio quality test on Linux using PulseAudio.
INSTALLATION
The test depends on PulseAudio virtual devices (null sinks). Without additional
arguments, run_audio_test.py expects a pair of sinks named "capture" and
"render". To create these devices at machine startup, place the provided
default.pa file in ~/.pulse. Alternately, the "pacmd" commands therein can be
run on the command-line to create the devices.
Similarly, place the provided daemon.conf file in ~/.pulse to use high quality
resampling in PulseAudio. This will reduce the resampling impact on the outcome
of the test.
Build all WebRTC targets as usual (or just the audio_e2e_harness target) to
generate the VoiceEngine harness.
USAGE
Run run_audio_test.py to start. The script has reasonable defaults and will
use the expected location of audio_e2e_harness. Some settings will usually
be provided by the user, particularly the comparison tool command-line and
regular expression to extract the quality metric.
An example command-line, run from trunk/
tools/e2e_quality/audio/run_audio_test.py \
--input=data/voice_engine/audio_short16.pcm --output=e2e_audio_out.pcm \
--codec=L16 --compare="comparison-tool" --regexp="(\d\.\d{3})"