Testing software tools with synthesized deviant voices for medicolegal assessment of occupational dysphonia

The measure of acoustic parameters related to dysphonia has long been one of the main research challenges in voice analysis. When the origin of the voice disorder is occupational, the medico-legal and insurance aspects must be considered, involving, when relevant, the definition of a percentage of physical impairment (usually 1-10% according to the guidelines). An objective aid capable of quantifying these levels is desirable for assisting the medico-legal expert. Vocal fold nodules in voice professionals are mentioned in the European List of Occupational Diseases. Generally, such voices are slightly to moderately deviant, and mainly characterized by audible air escape (breathiness), due to insufficient vocal fold closure. Mild amounts of jitter may also be present. No studies have been published so far testing the ability of software tools to discriminate levels of added noise in signals when the amount of noise is exactly known. In the present study, four program tools (BioVoice, PRAAT, MDVP and AMPEX) are tested on realistic synthesized voice signals corrupted by 4 slightly increasing levels of jitter and 10 levels of additive noise close to the degree of dysphonia occurring in real patients with vocal fold nodules. The results show that the four tools are able to correctly estimate both jitter and noise levels. Specifically for noise, in some cases the agreement is close to 100%. Thus they could be of help to clinicians in determining the level of impairment to quantify compensation in patients affected by an occupational voice disorder.