Skip to main content
. 2021 Mar 22;8(21):16035–16046. doi: 10.1109/JIOT.2021.3067605

TABLE II. LLDs for ComParE Feature Set. RASTA: Relative Spectral Transform; HNR: Harmonics to Noise Ratio; RMSE: Root Mean-Square Energy; and SHS: Subharmonic Summation. Details Can be Found in [55].

55 Spectral LLDs Group
MFCCs 1–14 Cepstral
Psychoacoustic sharpness, harmonicity Spectral
RASTA-filtered auditory spectral bands 1–26 (0–8kHz) Spectral
Spectral energy 250–650Hz, 1k–4kHz Spectral
Spectral flux, centroid, entropy, slope Spectral
Spectral roll-off point 0.25, 0.5, 0.75, 0.9 Spectral
Spectral variance, skewness, kurtosis Spectral
6 Voicing related LLDs Group
Inline graphic (SHS and Viterbi smoothing) Prosodic
Probability of voicing Voice Quality
log HNR, jitter (local and Inline graphic), shimmer (local) Voice Quality
4 Energy related LLDs Group
RMSE, zero-crossing rate Prosodic
Sum of auditory spectrum (loudness) Prosodic
Sum of RASTA-filtered auditory spectrum Prosodic