INVESTIGATION OF PITCH AND NOISE FEATURES EXTRACTED FROM VOICE SAMPLES OF HEALTHY AND PARKINSON AFFECTED PEOPLE USING STATISTICAL TESTS

Abstract

Parkinson Disease (PD) is characterized by reduced muscular movements, which also affects the speech production muscles. This leads to the degradation of speech quality and intelligibility, which is captured by extracting voice parameters like pitch, amplitude and noise features. The statistical comparison of these parameters with those of healthy subjects is a first step, which helps in the diagnosis of the disease. In this paper, different pitch and noise features are investigated. The investigated pitch features are mean, maximum and minimum pitch. The noise features analysed are Harmonic to Noise Ratio (HNR), Noise to Harmonic Ratio (NHR) and Glottal to Noise Excitation (GNE) ratio. These features extracted from both the PD affected and healthy people are then investigated to determine whether a statistically significant difference exists between the two groups using statistical tests, which include the descriptive statistics, correlation analysis, box-plots, Shapiro-Wilk test and Mann-Whitney U tests. The results obtained from the statistical tests performed on the speech signal features extracted from the speech samples shows that a statistically significant difference between the two groups is observed in all the pitch features, NHR and HNR features but not in the GNE feature.

Keywords

• Noise features
• Parkinson disease
• Pitch
• Statistical tests