Document Type : Original Article
Author
Azadi Square, University of Isfahan
Abstract
This study acoustically examines voice quality parameters in two groups of Persian-speaking men and women. The study aims to assess the ability of voice quality parameters to differentiate Persian speakers and to evaluate the extent to which these parameters capture speaker-specific information. Additionally, this research seeks to expand existing knowledge in the field of voice quality and address the limited scope of previous studies on Persian. Acoustic data were collected from 20 female and 20 male speakers in a laboratory setting. Multivariate analysis of variance (MANOVA) was used to analyze inter-speaker differences, and the Random Forest algorithm was employed to assess feature importance. Six voice quality parameters were selected for analysis: jitter (frequency perturbation), shimmer (amplitude perturbation), harmonic-to-noise ratio (HNR), the ratio of the amplitudes of the first and second harmonics (H1-H2), cepstral peak prominence (CPP), and fundamental frequency (F0). The results demonstrated significant acoustic differences among Persian speakers based on voice quality features, though the discriminative power of these features was not uniform. For male speakers, CPP, HNR, and H1-H2 were identified as the most discriminative features, respectively. For female speakers, F0, CPP, and HNR emerged as the key features for speaker identification. The findings highlight the significant role of voice quality parameters in identifying Persian speakers. However, achieving higher accuracy in speaker recognition systems requires considering gender differences and the relative importance of various variables. Moreover, the limited number of participants may affect the generalizability of the results. Thus, future studies are recommended to include larger and more diverse speaker samples.
Keywords
Main Subjects
)
Send comment about this article