PSQM+

PSQM+ was developed by KPN Research in 1997 [13]. The performance of PSQM+ is an improvement over that of the P.861 (PSQM) for loud distortions and temporal clipping by some simple modifications to the cognition module. PSQM+ can be applied to a wider range of distortions as an objective measure than PSQM. PSQM+ uses the same perceptual transformation module as PSQM. Similar to PSQM, PSQM+ transforms the speech signal into the modified loudness domain, and does not include temporal or spectral masking in its calculation of loudness. In order to improve the performance for the loud distortions like temporal clipping, an additional scaling factor is introduced when the overall distortion is calculated. This scaling factor makes the overall distortion proportional to the amount of temporal clipping distortion. Otherwise, the cognition module is the same as PSQM. PSQM+ adopts a simple algorithm in the cognition module to improve the performance of PSQM. The poor performance of PSQM for distortions like temporal clipping is due to the procedure calculating a scaling factor. The scaling factors are determined by the ratio of the energy of the distorted speech and the reference speech. This scaling factor scheme of PSQM works very well when a distortion results in additional energy. However, if a distortion results in reduced energy such as temporal clipping, which removes some of the signal energy, the estimate of distortion is proportionally smaller, and PSQM underestimates the actual distortion. Therefore, PSQM+ uses a simple modification to adopt an additional scaling factor to compensate for this effect. PSQM+ resolves one performance issue of PSQM on distortions such as temporal clipping. However, the performance of PSQM+ may be questionable for other different types of distortions.