Summary

The Chinese (CN) reference speech sample in ITU-T Rec. P.501 (04/2025) Annex D led to considerably lower scores for G.711 encoding (approximately 0.3 MOS lower on average) than the other reference speech samples in Annex D, when evaluated with ITU-T P.863 NB (narrowband mode).

After some analysis, it was observed that the reason for the low score was due to the female sentence characteristics. Applying a high-pass filter with a cut-off frequency of 120 Hz to the female sentence moved the combined female-male Chinese reference sample closer to the average of the P.501 Annex D samples.

Corrected wave files in all bandwidths are replaced in the ITU-T Test Signal Database for Annex D of ITU-T P.501.