Table 1.
Summaries of features.
| Feature Name | Num | |
|---|---|---|
| F0 | Vocal cord vibration cycle | 1 |
| VUV | Vector containing the binary voicing decisions | 1 |
| NAQ | Normalized amplitude quotient is presented as a method to parametrize the glottal closing phase | 1 |
| QOQ | The quasiopen period describes the duration of the glottal flow above 50% of the peak amplitude | 1 |
| H1H2 | Difference in glottal harmonic amplitude | 1 |
| PSP | Parabolic spectral parameter fitting a parabolic function to the low-frequency part of the estimated glottal flow | 1 |
| MDQ | The Maxima Dispersion Quotient, is proposed for discriminating breathy to tense voice | 1 |
| Peak Slope | Slope coefficient of a regression line fit to local peak by using wavelet analysis. | 1 |
| Rd | The Rd. shape parameter of the Liljencrants-Fant (LF) glottal model using the Mean Squared Phase (MSP) method based on MSPD2 | 1 |
| Rd-conf | 1 | |
| Creak | Detect creaky voice using acoustic features by an artificial neural network | 1 |
| MCEP | Transform the spectrogram into a Mel spectrum through the Mel scale filter bank, and then perform cepstrum analysis | 25 |
| HMPDM | Harmonic Model Phase Distortion Mean and Harmonic Model Phase Distortion Deviation are flexible representation of the glottal source based on the short-term statistics of the phase distortion | 25 |
| HMPDD | 13 | |
| Peak-to-RMS | Peak-to-RMS measure reflecting a local loudness metric related to waveform shape across a few pitch periods | 1 |
| Formant | Formants refer to areas where energy is relatively concentrated in the sound spectrum | 5 |
| MFCC-deltas | Reflecting the dynamic information of the spectrum envelope on a frame of voice signal | 20 |
| MFCC-delta-deltas | 20 |