Real-time visual displays for voice tuition

Abstract

The human voice is used regularly in day-to-day activity, particularly by teachers, politicians and singers. This can place a level of strain on their vocal mechanism. To counter this, professional voice users may undergo some form of vocal tuition. However, there is a prevalence of subjective, sometimes ambiguous tuition techniques.Vocal damage can therefore occur in spite of, or partially because of, the tuition techniques that are supposed to help the subject.It is hypothesised that a computer system which can a) objectively measure a large set of parameters relating to vocal performance, b) appropriately translate these parameters to a smaller set of easily interpreted measures, and c) dynamically display this information in an appropriate graphical form may be able to provide a measure of progress to complement instructive techniques during vocal tuition.A highly flexible, easy-to-use tool for the analysis and visualization of voice source and acoustic signals has been developed called ALBERT (Acoustic and Laryngeal Biofeedback Enhancement in Real-Time). The system can be used to provide real-time feedback on the state of a number of voice parameters. The visualization of these parameters is highly configurable, and includes 1D, 2D and 3D arrangements with optional colour mapping. ALBERT is novel in its ability to display many popular laryngeal and acoustic parameters in many informative graphical configurations. The system can be used in real-time to function as a tool for use by a voice tutor.To test the hypothesis, three studies are presented of vocally developing male and female subjects as follows: i) a six week course of vocal expression, ii) a group of actors over an eight month period, and iii) a group of singers over a time span of 2.7 years. Two voice parameters in particular have been identified as indicative of vocal performance: the closed quotient of the vocal folds (CQ) and the spectral amplitude in the frequency band within which the phenomenon of the 'singer's formant' is known to occur relative to the spectral amplitude of the complete acoustic output (Ratio). ALBERT has been used as a tool to provide real-time visual feedback of these parameters in the context of singing tuition. Results suggest that ALBERT is a tool with considerable potential for use in a voice tuition process. 2.4 General arrangement of the vocal organs (From Crystal, 1987) .. 11 2.5 Physiology of the human head (From Sataloff, 1992 2.10 Vocal tract profiles for two vowels. The shape is determined by the lip and jaw openings, the tongue shape, the velum, and the larynx.(After Fant, 1960;in Sundberg, 1987) ................ 262.11 The pattern of vocal fold movement (From Sundberg, 1987) .... 272.12 Idealised glottal area, flow and spectrum (After Pickett, 1991) ... 292.13 Spectra of simple waveforms (From Dodge & Jerse, 1985) ..... 302.14 Average source spectra from untrained voices (filled circles) and pro- 4.14 Pattern of change of mean CQ/Ratio for all male subjects, for all five assessment tasks. Spli...