Computer synthesis of disordered voices

Version 0.1.1 of SimuVox is available for download.

In this project, we develop a synthesizer of speech sounds (SimuVox) that simulates the timbre of disordered voices, i.e., voices that are perceived as abnormal often by consequence of a laryngeal pathology or a physiological dysfunction. We follow a physics-based strategy using models of the vocal fold vibration, glottal aerodynamics and acoustic wave propagation in the vocal tract.

Voice simulator
SimuVox 0.1.1

Physics-based synthesis allows for the control of the generated sound in direct terms of physiological parameters. SimuVox includes a lumped description of vocal fold tissue mechanics which is capable of simulating right/left fold asymmetries. It has an articulatory model for the vocal tract which controls its shape with articulatory parameters: jaw position, tongue body position and shape, tongue tip position, lip opening and protrusion, and larynx height. It also includes representations of the trachea and bronchi, nasal tract and paranasal sinuses, models of muscle jitter, neurological and physiological tremor, and airflow noise. Perceptual tests on rough, breathy and strained voices have shown that listeners confuse simulated and recorded voices with misclassifications errors around 40%, proving a good quality of the synthesis.

🔊 Audio

SimuVox may find applications for the investigation of physiological conditions that cause a particular abnormal vocal quality, for providing controllable synthetic stimuli in perceptual tests, and as a computational tool for training of clinicians, clinical diagnosis and treatment.

Some technical information on the synthesizer may be found in: J. C. Lucero, J. Schoentgen & M. Behlau. “Physics-based synthesis of disordered voices”, Interspeech 2013 (Lyon, France, 2013). [PDF]

Grants and collaborations

  • CNPq grant, with Mara Behlau, Center for Voice Studies (CEV) and Federal University of São Paulo, and Glaucya Madazio, CEV, 2015- 2018.

  • International Cooperation CNPq-FRS/FNRS (Belgium), with Jean Schoentgen, Université Libre de Bruxelles, 2012-2014.

Graduate students

  • Marina T. Englert, Ph.D. student, Federal University of São Paulo (Master in Science, 2016). Honorable Mention at the Best Student Award by The Voice Foundation (USA), 2017.