The modified LF voice source model of KLSYN88 was investigated through spectral and glottal waveform analysis of synthesized vowels, where each of the four parameters of the model was systematically varied; open quotient (OQ), speed quotient (SQ), spectral tilt (TL), and the amplitude of voicing (AV). There was particular emphasis on comparison with the original LF model. The spectral analysis indicates that the effects of changing parameter settings are broadly in agreement with expectations. However, results for TL showed that attenuation is less than expected and there was an indirect effect on the excitation strength (EE). Pulse skew turned out to be always higher than expected from SQ, which was also found to vary with changes to other parameters. Furthermore, as the LF parameter EE is affected by the settings of not only AV, but also SQ and TL, the mapping from EE to AV is not straightforward.
[1]
Ailbhe Ní Chasaide,et al.
PERCEPTUAL CORRELATES OF SOURCE PARAMETERS IN BREATHY VOICE
,
1999
.
[2]
D. Klatt,et al.
Analysis, synthesis, and perception of voice quality variations among female and male talkers.
,
1990,
The Journal of the Acoustical Society of America.
[3]
G. Fant.
Dept. for Speech, Music and Hearing Quarterly Progress and Status Report the Lf-model Revisited. Transformations and Frequency Domain Analysis the Lf-model Revisited. Transformations and Frequency Domain Analysis*
,
2022
.
[4]
Dennis H. Klatt,et al.
Software for a cascade/parallel formant synthesizer
,
1980
.