Measurements and Computation of the Loudness of Noises

Abstract

Computing the total loudness of noises by summing the loudness of octave bands, according to the method used by Mintz and Tyzzer, still gives results that deviate from subjective measurements. Apparently, part of this deviation is caused by masking in the lower octaves. There are two possible ways of taking account of masking, either by correcting the loudness of the single octave bands before summation—the corrections depend on the distribution of energy in the whole spectrum—or by dividing the spectrum into bands of such width that the contribution of the computed loudness of each band corresponds to the real effective magnitude. The two possibilities are investigated empirically; only the second is suitable as the basis for constructing an objective loudness meter. Using six bands for the summation of loudness proves most favorable. The upper five bands are nearly equal in width on the “mel”-scale, whereas the lowest band is of greater width. The results of subjective and objective measurements of different noises are discussed and compared with the results of loudness computations.