Formant (F1, F2) and VISC Features of Japanese Whispered and Normal Vowels

This dataset contains the acoustic measurements of Japanese whispered and normal (modal) vowels, extracted from the JVS (Japanese Versatile Speech) Corpus. The dataset includes the first two formants (F1, F2) and two VISC (Vowel-Inherent Spectral Change) features (VL and alpha) across 30 Japanese speakers, covering a total of 1,206 vowel tokens per speech phonation. Each CSV file contains the following 7 columns: vowel: The vowel category (e.g., a, i, u, e, o). speaker: Numeric speaker identifier (e.g., 1, 2, ..., 30), corresponding to the speaker numbers in the JVS corpus without leading zeros (e.g., 1 refers to jvs001). gender: Speaker gender ('M' for male, 'F' for female). f1: The first formant frequency in Hz. f2: The second formant frequency in Hz. vl: Vector Length, representing the magnitude of the dynamic spectral change (VISC feature). alpha: Trajectory angle, representing the direction of spectral movement in the F1-F2 space (VISC feature). The Original JVS Corpus: Takamichi S, Mitsui K, Saito Y, Koriyama T, Tanji N, Saruwatari H. JVS corpus: free Japanese multi-speaker voice corpus. arXiv:1908.06248