Acoustic and airflow spectral analysis of voice tremor

Acoustic spectral analysis has been used to describe voice tremor with some success, but no feature distinguishing pathological from normal tremor has been clearly identified. To assist in monitoring voice tremor associated with neurological diseases, objective and quantifiable measures that can distinguish between normal and pathological tremor are desired. This study explored the plausibility of using airflow and acoustic signals to quantify the frequency and amplitude of voice tremor and potentially to distinguish pathological from normal tremor. Subjects were 10 individuals with pathological tremor, most of them individuals with Parkinson's disease, and 10 gender and age-matched individuals with no voice disorder. Simultaneous acoustic and airflow signals were recorded during sustained vowel phonation. The acoustic intensity contours and the airflow signals were submitted to spectral analysis. A peak prominence ratio, defined as the ratio of the spectral peak energy to the overall signal energy, was calculated for each spectral peak below 30 Hz. For each subject, the 6 spectral peaks with the highest peak prominence ratios were selected. Frequency values of the 6 selected acoustic or airflow spectral peaks failed to distinguish tremor group from control group. Peak prominence ratios of the 6 selected acoustic spectral peaks were significantly higher for tremor group than for control group. Although spectral analysis of airflow signals was not useful in differentiating tremor group from control group, acoustic intensity contours and airflow time waveforms were highly and positively correlated in more tremor subjects (90%) than control subjects (40%). This finding suggests that the relationship between acoustic intensity contours and airflow time waveforms may reflect the presence and the source of voice tremor.