Features Explained

Rate of vocal fold vibration — perceived as pitch.

Mean Fundamental Frequency f0_mean_hz: Average rate of vocal fold vibration. Indicates typical pitch (age/gender differences).
Min / Max Fundamental Frequency f0_min_hz , f0_max_hz: Lowest and highest detected pitch; useful for range assessment.
Standard Deviation of f0 f0_std_hz: Pitch variability. High values may indicate instability or vocal tremor.
Pitch Frequency Range pfr_semitone: Range expressed in musical semitones. Narrow ranges can indicate pathology (paralysis, muscle tension).

Harmonics-to-Noise Ratio hnr_db: Ratio of periodic (harmonic) energy to noise, in dB. Lower HNR indicates breathy or hoarse voice. Normal > 20 dB; pathological < 15 dB
Harmonics-to-Noise Ratio (Voice Report) hnr_voice: Alternate HNR calculation using Praat's voice report algorithm.
Noise-to-Harmonics Ratio nhr: The inverse of HNR. Higher NHR = more aperiodic/noisy energy.
NHR (Harmonicity) nhr_harmonicity: NHR derived from the harmonicity (autocorrelation) method.
NHR (Voice Report) nhr_voice: NHR calculated via Praat's voice report.

Degree of Unvoiced Frames duv: Percentage of frames where pitch was not detected (voice stopped). High values may indicate severe voice issues.
Number of Voice Breaks nvb: Total count of phonation interruptions lasting longer than a few milliseconds.
Degree of Voice Breaks dvb: Total duration of all voice breaks combined, as a percentage of phonation time.

Cycle-to-cycle frequency variability — correlates with perceived roughness.

Local Jitter jitter_local: Average absolute difference between consecutive pitch periods, divided by average period. Normal < 0.5%; pathological > 1.0%
Local Absolute Jitter jitter_abs: Same as local jitter but expressed in microseconds (µs).
Relative Average Perturbation jitter_rap: Smoothed jitter averaging over 3 consecutive cycles — reduces random noise.
Pitch Perturbation Quotient jitter_ppq5: Jitter averaged over 5 consecutive cycles.
Difference of Differences jitter_ddp: Average absolute difference between consecutive differences of pitch periods — measures jitter "acceleration."

Cycle-to-cycle amplitude variability — correlates with perceived harshness or strain.

Local Shimmer shimmer_local_percent: Average absolute difference between consecutive peak amplitudes, divided by average amplitude. Normal < 3.0%; pathological > 5.0%
Local dB Shimmer shimmer_db: Amplitude perturbation expressed in decibels.
APQ3 shimmer_apq3: Smoothed shimmer averaging over 3 consecutive cycles.
APQ5 shimmer_apq5: Smoothed shimmer averaging over 5 consecutive cycles.
APQ11 shimmer_apq11: Smoothed shimmer averaging over 11 consecutive cycles.
Difference of Differences shimmer_dda: Average absolute difference between consecutive differences of amplitudes — shimmer "acceleration."

Maximum Phonation Time MPT: Longest sustained vowel duration (seconds). Indicates respiratory support and glottal closure efficiency. Normal > 15s (adults); > 10s (children)
Lowest Intensity ILow: Quietest dB level detected — useful for assessing dynamic range.
Number of Unvoiced Frames nuv: Raw count of silent/breathy frames in the recording.
Total Speech Segments seg: Number of detected speech chunks (separated by silent pauses).
Amplitude Variation vAm: Percentage of frames with rapid amplitude changes — marker of instability.
Frequency Variation vF0: Percentage of frames with rapid pitch changes — marker of instability.
Pitch Frames per_pitch: Total number of analysis frames where pitch was detected. Higher counts = more reliable analysis.
Number of Periods per_pp: Total number of individual glottal cycles counted.
Average Pitch Period avg_period_ms: Average time (ms) between consecutive glottal pulses (= 1000 / mean f0).

Analyze the overall frequency spectrum — capturing vocal quality, resonance, and vocal tract filtering.

Mean Cepstral Peak Prominence meanCPP: How prominent the pitch-related cepstral peak is. The single best acoustic marker of dysphonia severity. Normal > 12 dB; pathological < 8 dB
Std Dev of CPPS stdevCPP: Variability of cepstral peak prominence over time.
CPPS-derived F0 mean_cppF0: Average pitch derived from the cepstrum (vs. traditional autocorrelation).
Std Dev of CPPS F0 stdev_cppF0: Pitch variability from cepstral analysis.
Low-to-High Ratio meanLH_ratio: Energy ratio below vs. above ~1000 Hz. Higher = "dark"/muffled; lower = "bright"/thin. Useful for assessing hyponasality vs. hypernasality
Low-to-High Ratio (dB) meanLH_ratio_dB: Same ratio expressed in decibels.
Std Dev of LH Ratio stdevLH_ratio: How much the low-to-high ratio changes over time — high variability = poor resonance stability.
Std Dev of LH Ratio (dB) stdevLH_ratio_dB: LH ratio variability in decibels.
LTAS Slope mLTAS: Long-term average spectrum slope. Steeper = softer/breathier; flatter = louder/pressed.
LTAS Tilt tiltLTAS: Difference between low and high frequency slopes. Helps differentiate vocal fold stiffness from edema.

Mathematical combinations of raw acoustic parameters into single severity scores.

Dysphonia Severity Index DSI: Combines MPT, highest f0, lowest intensity, and jitter. Scale: +5 (healthy) to −5 (severe dysphonia)
Acoustic Voice Quality Index AVQI: Aggregates CPP, shimmer, HNR and other features. Scale: 0 (normal) to 10 (severely pathological)
Cepstral Spectral Index of Dysphonia CSID: Uses CPP and L/H ratio to quantify roughness and breathiness. Scale: 0 (normal) to 100 (severe)