سبك‌هاي گفتاري و شناسايي گوينده: نقش پارامترهاي كيفيت صدا

Speaking styles an‎d speaker identification: The role of voice quality parameters

زبان شناسي

Abstract This research was conducted with the aim of examining the role of speaking style (read an‎d spontaneous) in speaker identification an‎d determining the acoustic parameters affecting this process. In this study, 40 Persian speakers (20 men an‎d 20 women) participated an‎d produced their speech data in two speaking styles. Thirteen acoustic features related to voice quality were extracted based on the psychoacoustic model of Kreiman et al. (2014). Audio processing of the data was performed using Praat an‎d VoiceSauce software. Multivariate analysis of variance was used to examine the effect of speaking style an‎d speaker factor on acoustic features of voice quality. Additionally, multinomial logistic regression was employed to determine the relative contribution of each feature in explaining inter-speaker variations. The results of data analysis showed that in the male group, all features had significant differences in distinguishing between read an‎d spontaneous speaking styles. In the female group as well, all features except F3 an‎d F4 showed significant differences between the two read an‎d spontaneous speaking styles. Furthermore, the findings showed that all voice quality features had significant differences between speakers; however, their individual specificity varied depending on the speaking style an‎d gender of the speakers. Based on the results, in the male group, the features SHR, H1–H2, an‎d F0 respectively in read speaking style an‎d the features H1–H2, F0, an‎d CPP in spontaneous speaking style; an‎d in the female group, the features CPP, H2kHz–H5kHz, an‎d H4–H2kHz respectively in read speaking style an‎d the features F0, H2kHz–H5kHz, an‎d CPP in spontaneous speaking style had the greatest effectiveness in speaker identification. Overall, the findings indicate that although acoustic features of voice quality play a determining role in speaker identification, speaker identification systems should consider both speaking style an‎d gender in selec‎ting optimal acoustic features an‎d account for the effect of changing speech conditions on the value an‎d importance of parameters in the identity recognition process. Keywords: Speaker identification, Read speaking style, Spontaneous speaking style, Voice quality.

5 فصل

150805