Measuring Speech Quality across Mixed Networks

Olivier Willi   Thursday, 23rd June 2022

Users of VoLTE phones expect wideband speech. Of course, if the other end of the call is PSTN, 2G or 3G narrowband then their experience will not meet expectations. The planned or unexpected tendency for mixed networks to deliver narrowband instead of wideband speech can be verified with MultiDSLA.

POLQA enables speech quality measurements

Using ITU-T Rec. P.863POLQAenables speech quality measurements to be made on any network, from narrowband to super-wideband, with POLQA’s Super-wideband model. The score for narrowband speech will always be less than for wideband speech, given consistent impairments.

By way of demonstration, consider three transmission cases:

• Narrowband to narrowband
• Narrowband to wideband
• Narrowband to super-wideband

POLQA Super-wideband model

Using the POLQA Super-wideband model, the score should be similar for each case. We can apply a 48k sample rate signal, filtered to the IRS or MIRS narrowband characteristics, capture it at 8k, 16k and 48k sample rates and evaluate it with POLQA SWB.

The lower 8k measuring sample rate scores reflect the lack of any signal above 4kHz. The MIRS filtered files do have more energy above 4kHz than those that have been IRS filtered.

If we apply the 16k SWB signal and use a measuring sample rate of 8k, 16k and 48k, we see the score is limited to 4.6.