Acoustic Analysis of Voice: A Tutorial

This tutorial reviews acoustic methods that have been used to characterize vocal function. The most persuasive argument for the use of acoustic measures is that all of the information used by listeners to make judgments about speech is to be found in the acoustic signal. Acoustic methods have been used clinically to differentiate normal from abnormal voices, to aid in differential diagnosis, to evaluate the relative effectiveness of different treatment approaches, and to track progress in voice therapy. The measures discussed here focus on quantifying the degree of periodicity, the shape of the spectrum, and the range of vocal intensity. Introduction Fundamental to the logic underlying many acoustic measures of voice is the simple observation that the voice is produced to please the human ear. Listeners make both linguistic and aesthetic judgments about voice signals, and essentially all of the information that controls these judgments is embedded within the acoustic signal. It is also the case that the disturbances in glottal vibratory patterns that are associated with laryngeal pathologies will be reflected in changes in the acoustic signal that is radiated from the lips. Also, as has been noted by many investigators, the voice signal is perhaps the most noninvasive and easily obtained measure of vocal performance. As straightforward as this logic may seem, uncovering the fundamental relationships between acoustic properties and perceived vocal qualities, or the relationships between the characteristics of the acoustic signal and the underlying laryngeal vibratory patterns that shape the sound wave, has not proven to be an especially easy problem. Goals of Acoustic Analysis The major clinical applications of acoustic analysis tend to fall into three broad categories: (1) screening, (2) diagnostic support, (3) evaluation of the relative effectiveness of different treatment approaches, and (4) assessment of progress throughout the course of treatment (see Laver Hiller & Beck (1992) for discussion and a slightly expanded list of applications). Screening applications are often directed at rapid and inexpensive early detection of serious conditions such