Statistical analysis of F0 for Iraqi Arabic accent corpus

This paper presents a statistical analysis of the fundamental frequency (F0) in Iraqi-accented speech. It includes the mean, variance, mode, skewness, kurtosis, Probability Density Function (PDF), and Cumulative Distribution Function (CDF) of F0. The researchers created a corpus of speakers from near Baghdad, comprising 28 females and 19 males of varying ages, recorded at 16 kHz with 16-bit resolution. The corpus is uploaded to the researchers’ GitHub repository. For females, the average values of the RAPT, PRAAPT, SHR, and YIN algorithms are: mean =209 Hz, standard deviation =57.8 Hz, variance =4866 Hz2, mode =102.8 Hz, skewness =2.4, and kurtosis =13.81. For males, the average values are: mean =136 Hz, standard deviation =57.8 Hz, variance =3874 Hz2, mode =102.8 Hz, skewness =2.4, and kurtosis =13.81. The PDF and CDF of F0 are calculated and distributed. The results for the Iraqi accent show slight differences compared to those of the Saudi accent and to English TIMIT and formal Arabic. The research aims to create a corpus for a previously unexplored region of Iraq and to make statistical analysis available to other researchers in future studies. In Iraq, speech DSP researchers were suffering from a lack of such important datasets and their analysis.