Speech Quality Testing - PESQ

What is PESQ (Perceptual Evaluation of Speech Quality)

Perceptual evaluation of speech quality (PESQ) is a recognised method of accurately testing the level of quality that will be perceived by the user of a VoIP network. Successful measurement requires a WAV file to be generated by two parties connected at different network nodes, the files received at each end are recorded and compared with a reference file in order to determine any loss of quality en-route.

The P.862 ITU PESQ standard combines the excellent psycho-acoustic & cognitive model of PSQM+ with a time alignment algorithm to accurately manage the variation in delay. PESQ provides a score in the range of 1 to 5 where 1 is unacceptable & 5 is excellent. A typical range for VoIP is 3.5 to 4.2. WAV files for PESQ measurement are available from the ITU-T website in original (or) & degraded (dg) format to meet a variety of applications.

Typical PESQ Applications

  • Make several PESQ measurements during one long duration call in order to test speech quality over time as network demands fluctuate.
  • Provide P.862 ITU PESQ test results to equipment vendors, network operators & end users.
  • Test the effect when using different codec types.
  • Test and evaluate network equipment & services.
  • Ensure Quality of Service (Q.O.S).

PESQ Option for Zumara or Run-Time Software

  • Measure PESQ from LAN to LAN or from LAN to WAN through a Gateway.
  • The software injects a preceding pulse to ensure the exact start of each file is known for comparison. This ensures highly accurate results even when tests are made over International links.
  • The PESQ results table provides an indication of loss, jitter, PESQ score (P862), MOS score (P862.1), noise score & speech score amongst others.
  • PESQ measurements can be made between any two physical interfaces to provide total network visibility. This is important due to the number of devices and codec types that may be encountered by a call on route from A to B.

As well as the PESQ score expected from a perceptual evaluation of speech quality calculation, Zumara or Run-Time software, will provide a MOS score and other useful scores with level & noise measurements. This provides the user with a clear and comprehensive quality indication as shown in the example below.

back to top