Correlation between headphone and telephone-handset listener opinion scores for single-stimulus voice coder performance assessments

It has often been suggested that digital voice coder performance may depend on the type of listening set used in listener opinion assessments. In this letter, the quantitative impact of using headphones and telephone handsets is presented for single-stimulus MOS assessments. From this analysis, it is derived that there are significant numerical differences between tests employing these types of listening instruments and that cardinal differences in performance may be consistent and thus predictable. This study is extended to an ordinal analysis from which it is derived that although ranked performance does vary, depending on which of two listening instruments is employed, the two rankings are highly correlated.<>