Objective: To investigate and quantify the presence of noise in the vocal signals of Parkinson’s disease (PD) patients and healthy volunteers (HV) using acoustic measures, specifically focusing on GNE (Glottal-to-Noise Excitation ratio), HNR (Harmonics-to-Noise Ratio) and HFNO (High-Frequency Noise Energy).

Background: PD is a neurodegenerative disorder significantly impacts speech and voice production. One of the hallmark features of voice impairment in PD is the increased presence of noise in the vocal signal [1-3]. This feature is closely associated with hypokinetic dysarthria, a speech disorder commonly observed in PD.

Method: The study draws on data from Brainphone project (Russia). The speech recordings (sustained /a/ vowel) were sampled at 16 kHz with 32 bits resolution, collected over voice recorder in .wav format. All participants provided written informed consent.

Results: There were 109 unique patients with advanced PD (PwP) (M–39, F–79, H&Y stage 3–68, stage 4–41), 92 – healthy volunteers (HV) (male–33, female–59). Results on noise assessment metrics can be found in the Table 1 and Pic. 1-3.

A two-tailed t-test showed that the difference between HV and PwP with respect to HFNO was statistically significant (p = .015), the effect size is small (d = 0.35). The same is true for HNR, where the difference between HV and PwP was statistically significant (p = .003), the effect size is small (d = 0.43). For GNE a two-tailed t-test was also statistically significant (p = <.001) with the medium effect (d = 0.7).There’re no significant difference between the groups of the variable “Sex” in relation to the HFNO, HNR and GNE (p>0.1 in every case).

Logistic regression analysis was performed to examine the influence of HFNO, HNR & GNE to predict PD. It shows that the model as a whole is significant (Chi2(3) = 24.21, p <.001, n = 200). But GNE is the only measure that has statistically significant influence on the group prediction (b = 4.2, odds ration = 67). AUC for GNE is 0.711 (Pic. 4).

Conclusion: Research studies have consistently demonstrated that PD patients exhibit significant changes in GNE, HNR, and HFNO compared to healthy controls. These changes are attributed to the underlying pathophysiology of PD, including bradykinesia (slowness of movement), rigidity, and reduced coordination of the vocal folds. In our study we showed the GNE acoustic measure is a significant feature in analyzing voice characteristics in PD.

Figure 1. HFNO

Figure 2. HNR

Figure 3. GNE

Figure 4. AUC for GNE

Table 1. Descriptive characteristics

References: 1. Bang, Y., Min, K., Sohn, Y., & Cho, S. (2013). Acoustic characteristics of vowel sounds in patients with Parkinson disease.. NeuroRehabilitation, 32 3, 649-54 . https://doi.org/10.3233/NRE-130887.
2. Shen, J., Zhang, T., Huang, F., Zhou, H., Teng, F., Hakyung, K., & Jin, L. (2019). Study of voice disorder based on acoustic assessment in Parkinson ′s disease. , 52, 613-619. https://doi.org/10.3760/CMA.J.ISSN.1006-7876.2019.08.003.
3. Roland, V., Huet, K., Harmegnies, B., Piccaluga, M., Verhaegen, C., & Delvaux, V. (2023). Vowel production: a potential speech biomarker for early detection of dysarthria in Parkinson’s disease. Frontiers in Psychology, 14. https://doi.org/10.3389/fpsyg.2023.1129830.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/noise-assessment-acoustic-measures-of-sustained-vowel-in-patients-with-parkinsons-disease/