On the Relationship between Glottal Pulse Shape and Its Spectrum: Correlations of Open Quotient, Pulse Skew and Peak Flow with Source Harmonic Amplitudes

Abstract

This paper explores the relationship between the glottal pulse amplitude (Up) and the amplitude of the first harmonic (H1), as well as the combined effects of Up, the open quotient (Oq) and degree of pulse asymmetry/skew (Rk) on the low end of the source spectrum. This serves to elucidate their relationship to the H1-H2 estimate, widely used to make inferences on changes in Oq and voice quality. It has been suggested that H1 is mainly determined by Up and that the pulse shape has a relatively small impact. To investigate this, a series of glottal pulses were generated using the LF model, where Up was kept constant, while Oq and Rk were systematically varied. The resulting harmonic amplitudes of these pulses show that Up is not the sole determinant of H1. Rather, H1 is highly dependent on Oq and to a certain degree also on Rk. Although the effects of these parameters on the lowest harmonics is rather complex, we find that the H1-H2 measure is broadly correlated with Oq. However, there is also a strong effect of differences in glottal skew, particularly at high Oq values, which could invalidate inferences on Oq and voice quality from estimates of H1-H2.

Time to Frequency Domain Mapping of the Voice Source: The Influence of Open Quotient and Glottal Skew on the Low End of the Source Spectrum

“…However, a re-examination of the LF model simulations in [26] lead us to reconsider these findings, and motivate the present study. As illustrated in Figure 1, an LF pulse with very high Oq (0.85) and high skew (low Rk of 0.15) yields an effective Oq that is considerably lower than the numerical value of the model (0.5 rather than 0.85).…”

“…In [26] we used the LF voice source model to analyse the relationships between the glottal pulse shape and the amplitudes of source spectral components. However, for reasons outlined above, we here carry out essentially the same analysis, but instead using a different source model.…”

“…We also reanalysed the data from our earlier study [26] using OQe and RKe for the LF model (instead of Oq and Rk) for the correlation with H1 and H2 amplitudes. The F-model (see Figure 1, right panel) is determined by the two expressions in (1), which generate the glottal pulse of the opening phase and closing phase respectively.…”

“…Ideally, one would wish to be able to map reliably between the time and frequency domains to characterise the source (see, for example [21,13,24,25,26]). In this paper, we build on an earlier study [26], which focussed on the low end of the source spectrum, and which used LF model [27] simulations to explore the correlation of time domain parameters with the levels of the first three harmonics. The time domain parameters considered included the peak flow (Up) -described by Fant and Lin [24] as the principle determinant of H1 -and the pulse shape parameters open quotient (Oq) and glottal pulse skew (Rk).…”

“…As illustrated in Figure 1, an LF pulse with very high Oq (0.85) and high skew (low Rk of 0.15) yields an effective Oq that is considerably lower than the numerical value of the model (0.5 rather than 0.85). This is due to the fact that during the initial part of the opening phase of the LF pulse, the flow is very close to zero -a consequence of the exponentially [26][27][28][29]).…”