POLQA and PESQ: definition and a practical example
Short answer: POLQA (ITU-T P.863) and PESQ (ITU-T P.862) are ITU standards that calculate a MOS-LQO score by comparing the audio received on a call with the reference audio.
In Portuguese: POLQA e PESQ
What are POLQA and PESQ?
POLQA and PESQ are algorithms standardized by the ITU to calculate a call’s voice quality from its audio. Both are full-reference: they compare the audio that arrived at the other end with the original audio and return a MOS-LQO score, on the 1-to-5 MOS scale.
PESQ (Perceptual Evaluation of Speech Quality) is ITU-T P.862, created for narrowband speech, classic 8 kHz telephony. For many years it was the industry reference for measuring voice quality objectively. On January 5, 2024 the ITU withdrew P.862 and began pointing to P.863 in its place.
POLQA (Perceptual Objective Listening Quality Analysis) is ITU-T P.863, the successor. It also covers wideband and super-wideband speech, which matters for modern codecs, and handles distortions that PESQ assessed poorly. POLQA is distributed under a commercial license.
Both rely on a model of how the human ear perceives distortion. The score does not say what went wrong; it says how far the received sound drifted from the original to the ears of an average listener.
In practice
| Algorithm | Source | Status | Access |
|---|---|---|---|
| PESQ | ITU-T P.862 | Withdrawn on 1/5/2024 | Older implementations still around |
| POLQA | ITU-T P.863 | Standard in force | Commercial license |
| ViSQOL | Google Research | Maintained, version v3 | Open source |
An example of why scores should not be mixed: a 2019 report shows a MOS of 3.9 calculated with PESQ, and a test today shows 3.6 with another algorithm. The 0.3 drop may mean nothing, because each algorithm has its own mapping and its own ceiling. To track a line over time, the algorithm and the mode must always be the same.
How Monitor uses this
Monitor uses neither POLQA nor PESQ. It measures MOS-LQO with ViSQOL, in speech mode, comparing the audio received on each answered test call with the circuit’s reference recording. The reason to explain all three options here is that many scores on the market come from PESQ or POLQA. See voice quality with MOS-LQO.
Frequently asked questions
Can PESQ still be used?
PESQ implementations are still around, and many older reports use its scores. But the ITU withdrew ITU-T P.862 on January 5, 2024 and points to P.863, POLQA, as its replacement. For a new process, PESQ is no longer the standard in force, and its scores should not be compared directly with POLQA's.
Do POLQA and ViSQOL give the same score?
Not necessarily. Both compare the received audio with the reference and return a 1-to-5 MOS-LQO, but they use different perception models and different mappings to the scale. The same call can get different scores. Always compare scores from the same algorithm and the same mode.