Comparison of multiple voice source parameters in different phonation types

Abstract

A large sample of vowels produced by male and female speakers were inverse filtered and parameterized using 21 different glottal flow parameters. The performance of the different parameters in expression of the phonation type was then tested using objective statistical methods. The comparison of the results revealed marked differences in the parameters' performance, and therefore, guidelines for parameter use and comparison were established.

Analysis and classification of phonation types in speech and singing voice

“…Different phonation types such as whisper, breathy, tense, creak and falsetto can be produced by changing the activation of the laryngeal muscles [2,8,15]. Breathy and tense are often considered to be the two opposite ends of the voice quality continuum [3,16]. In this study, three phonation types of speech (breathy, modal and tense) from the continuum are considered as subjects of interest.…”

“…Variations in the vibration mode of the vocal folds described above result in differences in the shape of the glottal flow pulse between phonation types. The glottal flow waveform varies from a smooth, almost symmetric pulse in breathy phonation to an asymmetric pulse with sharp edges in tense phonation [16,29]. This kind of time domain variation is reflected in the decay of the spectral envelope of the glottal source in the frequency domain [30,31].…”

“…This kind of time domain variation is reflected in the decay of the spectral envelope of the glottal source in the frequency domain [30,31]. By taking advantage of both the time and frequency domain, many glottal source features have been developed to discriminate phonation types using flow waveforms estimated by glottal inverse filtering (GIF) [16,3,32]. Time domain features (such as the open quotient, the quasi-open quotient (QOQ), the closing quotient (CQ) and the speed quotient (SQ)), and amplitude-based features (such as the normalized amplitude quotient (NAQ)) have been widely used to parameterize the glottal flow and its derivative [16,28,33].…”

“…By taking advantage of both the time and frequency domain, many glottal source features have been developed to discriminate phonation types using flow waveforms estimated by glottal inverse filtering (GIF) [16,3,32]. Time domain features (such as the open quotient, the quasi-open quotient (QOQ), the closing quotient (CQ) and the speed quotient (SQ)), and amplitude-based features (such as the normalized amplitude quotient (NAQ)) have been widely used to parameterize the glottal flow and its derivative [16,28,33]. Examples of frequency domain features are the amplitude difference between the first and second harmonic (H1-H2) [31], the harmonic richness factor (HRF) [8] and the parabolic spectral parameter (PSP) [34], which measure the decay of the glottal flow spectrum.…”

Excitation-based Voice Quality Analysis and Modification

“…In previous works [9], [10], Alku et al proposed the Normalized Amplitude Quotient and the Quasi-Open Quotient as two efficient time-domain parameters characterizing respectively the closing and open phase of the glottal flow. These parameters are extracted using the Aparat toolkit [11] from the glottal source estimated here by the complex cepstrum .…”