Synthesis of disordered voices

The context of this project is the clinical assessment of voice. Auditory and acoustic assessment of voice (and by extension speech) is to laryngology and speech therapy what electrocardiography is to cardiology and electroencephalography to neurology. That is, it reports the function of the laryngeal oscillator and the adequacy of the produced vocal timbre by relying on methods of investigation that are not intrusive and do not obstruct the patient’s production of speech. Synthetic speech contributes to that purpose as a computational tool to facilitate tests and training, and also to explore and understand the genesis of abnormal vocal qualities.

Vocal tract

A disordered voice is a voice that is perceived as abnormal with regard to pitch, loudness or timbre, and is often the consequence of a laryngeal pathology or some physiological dysfunction. In this project, we develop synthesizeres of speech sounds which are capable of simulating the timbre of disordered voices with an acceptable level of naturalness. We follow a physics-based strategy using models of the vocal fold vibration, glottal airflow and acoustic wave propagation in the vocal tract.

Listen to some voice samples produced by the synthesizer. Its most recent version is described in our poster in the 9th International Conference in Voice Physiology and Biomechanics (2014), avaliable on the same link.

Colaborations

Dr. Jean Schoentgen, Université Libre de Bruxelles, Belgium.
Dr. Mara Behlau, Center for Voice Studies and UNIFESP, Sao Paulo.
Dr. Glaucya Madazio, Center for Voice Studies, Sao Paulo.

Students

Marina T. Englert, Master's course in Disorders of Human Communication, UNIFESP.
Elias Amadeus, Bachelor's course in Computer Science, University of Brasilia.

Support

CNPq (Brazil)