| To investigate the impact of long-time spans on speakers' acoustic characteristics, two data groups with different time spans were established in this paper. The experimental group spanned 16 years (the speakers were 19-20 years old when first recorded in 2005, and 35-36 years old when recorded again in 2021), while the control group spanned 3 months (with the speakers being 19-20 years old). Both groups included the Mandarin reading speeches of 10 female native speakers from Beijing. The extracted acoustic parameters encompassed the first four formants of the seven monophthongs (/a/, /?/, /u/, /i/, /y/, /?/, /?/), the Long-term Fundamental Frequency (LTF0) and the (first four) Long-term Formant (LTF) distribution of the corpus. Non-parametric tests revealed that the mean value of LTF0 in the experimental group decreased significantly over 16 years (in contrast, no significant change was observed over 3 months in the control group). Most formant values for the monophthongs exhibited no significant differences, and the vast majority of the LTF parameters showed no significant changes. Linear discriminant analysis indicated that, under the same circumstances, only when LTF0 was included as a predictor variable, would the discriminant rate of the control group be notably higher than that of the experimental group. Consequently, it can be inferred that, for female speakers within the youth period (19-40 years old), the primary effect of the 16-year-time span on the acoustic parameters can be summarized as follows: the mean value of LTF0 decreases significantly, while most formant parameters remain relatively stable. This conclusion is of great value for the forensic voice comparison. |
