An overview of issues in quantification of the tonalness of sounds

Abstract

Audible pitches in noise increase annoyance, and tone corrections to noise level metrics have been used in noise impact quantification for over 50 years. These corrections are a function of the prominence of the tone relative to the nearby spectral components. Frequency masking, the width of the tonal feature, and the frequency may also be taken into account. All measures are derived from spectral estimates. When a single or well isolated tonal components are present in noise, sounds are stationary and sufficient data is available for good spectral estimation, many tonalness metrics (Prominence Ratio, Tone to Noise Ratio, Tone Audibility, and Aures Tonality) work well. More challenging is quantification of tonalness when: the frequency is varying with time, either slowly or quickly; there is a harmonic series that is predominantly heard holistically not as a series of individual pitches; or there is beating because of the presence of multiple tones within a critical band. In music, the characteristics of harmonic families help us identify different instruments and manipulation of harmonic series in machinery noise can make tonal sounds more or less pleasant. The challenges of quantifying tonal contributions and their impact in these more complicated cases will be described