Description
A standardised listening test in which trained listeners rate several audio versions against a known reference, with a hidden reference and low-quality anchors to validate ratings.
MUSHRA asks listeners to rate multiple stimuli on a 0-100 continuous quality scale, simultaneously, with free switching between them and an open reference. Among the stimuli are a hidden copy of the reference and one or more anchors (e.g. a 3.5 kHz low-pass filtered version). Listeners who fail to rate the hidden reference near 100 are excluded under post-screening rules. The design yields sensitive, reliable comparisons for intermediate audio quality. For UX, it is a model of rigour for any perceptual comparison: voice assistant TTS, notification sounds, call quality, audio in video products. Its anchors and hidden reference make rater reliability measurable.
Attributes
- Discipline
- Neuro, Biometrics & Affect
- Loss type
- Unmigrated
- Stage
- Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Prepare stimuli: the reference, the systems under test, a hidden reference and the standard anchors, loudness-normalised.
- Recruit and train listeners (typically 15-20 experienced listeners) with a familiarisation phase.
- Run trials with all stimuli for an item presented together in random order, rated on 0-100.
- Apply post-screening: exclude listeners rating the hidden reference below 90 for more than 15% of items.
- Analyse with mixed models or ANOVA, reporting means and 95% confidence intervals per system.
- Relate quality differences to perceptual attributes through follow-up descriptive analysis if needed.
Sources
- ITU-R BS.1534-3 (2015) Method for the subjective assessment of intermediate quality level of audio systems
- Bech & Zacharov, Perceptual Audio Evaluation: Theory, Method and Application (2006)
Also known as
- MUSHRA
- Multiple Stimuli with Hidden Reference and Anchor
- ITU-R BS.1534
Categories
- Analytics & Behaviour
- Neuro, Biometrics & Affect