Codec listening test
Most tests take the form of a double-blind comparison. Commonly used methods are known as "ABX" or "ABC/HR" or "MUSHRA". There are various software packages available for individuals to perform this type of testing themselves with minimal assistance.
In an ABX test, the listener has to identify an unknown sample X as being A or B, with A (usually the original) and B (usually the encoded version) available for reference. The outcome of a test must be statistically significant. This setup ensures that the listener is not biased by his/her expectations, and that the outcome is not likely to be the result of chance. If sample X cannot be determined reliably with a low p-value in a predetermined number of trials, then the null hypothesis cannot be rejected and it cannot be proved that there is a perceptible difference between samples A and B. This usually indicates that the encoded version will actually be transparent to the listener.
In an ABC/HR test, C is the original which is always available for reference. A and B are the original and the encoded version in randomized order. The listener must first distinguish the encoded version from the original (which is the Hidden Reference that the "HR" in ABC/HR stands for), prior to assigning a score as a subjective judgment of the quality. Different encoded versions can be compared against each other using these scores.
In MUSHRA (MUltiple Stimuli with Hidden Reference and Anchor), the listener is presented with the reference (labeled as such), a certain number of test samples, a hidden version of the reference and one or more anchors. The purpose of the anchor(s) is to make the scale be closer to an "absolute scale", making sure that minor artifacts are not rated as having very bad quality.
Many double-blind music listening tests have been carried out. The following table lists the results of several listening tests that have been published online. To obtain meaningful results, listening tests must compare codecs' performance at similar or identical bitrates, since the audio quality produced by any lossy encoder will be trivially improved by increasing the bitrate. If listeners cannot consistently distinguish a lossy encoder's output from the uncompressed original audio, then it may be concluded that the codec has achieved transparency.
Popular formats compared in these tests include MP3, AAC (and extensions), Vorbis, Musepack, and WMA. The RealAudio Gecko, ATRAC3, QDesign, and MP3pro formats appear in some tests, despite much lower adoption as of 2007[update]. Many encoder and decoder implementations (both proprietary and open source) exist for some formats, such as MP3, which is the oldest and best-known format still in widespread use today.
|Source||Dates||Formats||Bitrate (kbit/s)||Codecs||Musical genres||Samples||Listeners||Best Result||Comments|
||1||16||Musepack and AAC|
|ff123||October 2001-January 2002||multiple||~128||
||Various||3||25-28||Musepack or Vorbis|
||Various||12||24-41||MP3pro||Both Vorbis variants were a close second.|
|Roberto Amorim||June 2003||AAC||128 CBR||
|Roberto Amorim||July 2003||multiple||~128||
||Various||12||14-24||Musepack||AAC, WMA, and Vorbis tied for close second|
|Roberto Amorim||September 2003||multiple||~64||
||Various||12||30-43||Nero HE AAC||This test showed that listeners preferred 128 kbit/s MP3 audio encoded by LAME to all the tested codecs at 64 kbit/s, with greater than 99% confidence:
"No codec delivers the marketing plot [sic] of same quality as MP3 at half the bitrates."
|Roberto Amorim||January 2004||MP3||~128||
||Various||12||11-22||LAME||The author noted that the results may have been affected by the use of an outdated version of the Xing encoder and non-optimal settings for ITunes.|
|Roberto Amorim||February 2004||AAC||~128||
||Various||12||19-29||iTunes||Open-source FAAC codec improved greatly since previous test|
|Roberto Amorim||May 2004||multiple||~128||
||Various||18||12-27||aoTuV and Musepack|
|Roberto Amorim||June 2004||multiple||32 CBR||
|HydrogenAudio user "guruboolez"||July 2004||multiple||~175||
|HydrogenAudio user "guruboolez"||August 2005||multiple||~180||
||Classical||18||1||aoTuV||The author reflects on substantial improvements in Vorbis encoding since his previous test (above):
"Vorbis is now –thanks to Aoyumi [creator of aoTuV]– an excellent audio format for 180 kbit/s encodings (and classical music)."
|gURuBoOleZZ (French)||August 2005||multiple||~96||
||Classic, various||150 classical, 35 various||1||aoTuV and AAC tied (classical), aoTuV (various)||The author selected each participating encoder by pitting multiple encoders against one another in an initial "Darwinian phase." For example, LAME was chosen as the representative MP3 encoder because it clearly outperformed four other MP3 encoders on a subset of the full sample corpus.|
|Sebastian Mares||December 2005||multiple||~140 (nominal 128)||
||Various||18||18-30||4-way tie (all except Shine)||"I think this test shows that with the current encoders, the quality at 128 kbit/s is very good... It's time to move to bitrates like 96 kbit/s or even lower (64 kbit/s)."|
||Various||18||10-20||5-way tie (all except anchors)||"... it seems that overall, plain HE-AAC might be better than HE-AAC v2 at this bitrate, but a lot more samples would be needed to be able to draw definitive conclusions regarding this.|
|Sebastian Mares||November 2006||multiple||~48||
||Various||20||22-34||Nero Digital||WMA Professional and aoTuV tied for second|
|Sebastian Mares||July 2007||multiple||~64||
||Various||18||21-33||Nero Digital and WMA Professional|
|Sebastian Mares||October 2008||MP3||~128||
||Various||14||26-39||5-way tie (all except l3enc)||"The quality at 128 kbps is very good and MP3 encoders improved a lot since the last test." Also notes that Fraunhofer and Helix codecs are several times faster at encoding than LAME, although virtually identical in terms of perceived audio quality.|
|HydrogenAudio user IgorC (March/April 2011)||March 2011||multiple||~64||
||Various||30||25-13||CELT||In results, CELT is referred to as Opus, its name when later standardized.|
|HydrogenAudio user IgorC (July/August 2011)||July/August 2011||LC-AAC||~96||
|HydrogenAudio user "Kamedo2"||May 2013||MP3||~224||
||Various||25||1||4-way tie (all except BladeEnc low anchor)||Most impairment grades rated between 4 (perceptible but not annoying) and 5 (imperceptible). Both speech samples transparent (p<0.02) except for the low anchor.|
|HydrogenAudio user Kamedo2 (open at least until August 30th 2014)||July/August 2014||multiple||~96||