Signal-to-noise ratio as a predictor of speech transmission quality

Signal-to-noise ratio expressed in decibels may be generalized to the form
10 \log S^{a}/N^{b}
, where S and N are, respectively, signal and noise values expressed in power units. In the usual definition of signal-to-noise ratio, the exponents a and b both assume values of +1, that is, the ratio a/b = 1. This study was designed to examine the optimum value of a/b as a predictor of speech transmission quality using two sets of speech transmission quality data. Results of factor analysis, multiple regression analyses, and canonical correlation analyses showed that values of a/b greater than one were better predictors of transmission quality than a/b = 1.