A discussion on the origin and treatment of noise in industrial environments - Frequency weighting contours for predicting the speech interfering aspects of noise

Abstract

Physical and psychophysical schemes that purported to measure the speech interfering aspects of noise were examined in a series of papers by Klumpp & Webster (1962, 1963) and Webster & Klumpp (1963). Sixteen diverse-spectrum noises were adjusted in level so that listeners hearing monosyllabic (rhyme) words at a constant level of 78 dB from a loudspeaker obtained 50 % word intelligibility scores. Articulation index (a.i) calculations (see Kryter 1962 a ) predicted the speech-interfering properties of the noises very well. However, other, and simpler, schemes worked just as well; for example, speech interference level (s.i.l.) calculations (see Beranek 1954), based on octaves centred at 425, 850 and 1700 Hz, or 500, 1000 and 2000 Hz. The A-weighting and Din 3 networks (see Peterson & Bruel 1957), of a sound level meter (s.l.m.) were good, but the conventional use of noise criteria (n.c.), or alternate noise criteria (n.c.a.) (see Beranek 1957), curves did not work well unless (1) only that part of the curves centering on the octaves 500, 1000 and 2000 Hz was used, and (2) the noise spectra were allowed to ‘average through’ a contour and not just touch it at a peak value.