Perturbation Analysis of Synthetic Voice with Subharmonic Modulation
Objective: Subharmonic modulation is often observed in the voice signals of those with vocal fold lesions or vocal fold paralysis. The time-domain behaviors of these signals include cycle-to-cycle perturbations of amplitude, period, or shape. Cavalli and Hirson (1999) observed that the presence of the subharmonics correlated with the increases in amplitude and pitch perturbation measures among pathological voice samples. This study furthers our understanding of how the subharmonic modulation propagates through the voice production system. The effects of both amplitude and frequency modulations and their mixtures are investigated systematically with *in silico* voice synthesis using the kinematic vocal fold model coupled with the wave-reflection vocal tract model.
Methods: A symmetric kinematic vocal fold model is driven by a modulated sinusoidal lateral displacement function. Subharmonic modulation is introduced to the displacement function uniformly along both lateral and sagittal planes with five control parameters: modulation period, amplitude-modulation (AM) extent, AM phase, frequency-modulation (FM) extent, and FM phase. To limit the modulations to produce subharmonics, the modulation period is locked to an integer multiple of glottal cycle period. A male voice (Fo=101 Hz) with sustained /_/ is considered for the study. Four synthesized signals are observed: radiated acoustic pressure, glottal flow, contact area, and glottal area. Two perturbation measures (shimmer and jitter) and the subharmonics-to-noise ratio (SHR) are evaluated by a custom algorithm with known modulation properties.
Results: Key results include that jitter due to the FM vibration is consistently measured across four signals while shimmer due to the AM vibration varies across the signals. Both FM and AM vibrations yield an increase in both jitter and shimmer in the glottal flow and radiated pressure. All the measures are sensitive to the modulation phases with additional dependency on the modulation period.
Conclusions: This study demonstrates how different aspects of the subharmonic modulation of the vocal fold vibration affect the produced voice signals. The subharmonic characteristics observed at the source (HSV or EGG) do not translate directly to those of the resulting acoustic signals except for the modulation period.