Extract acoustic and MFCC features from voice recordings to characterize within-subject changes over time. Designed for observational studies and epilepsy voice-biomarker research.
A simple workflow for longitudinal voice research.
Voice sample via mobile or web app
Acoustic features & MFCCs
Within-subject reference
Statistical deviation
PDF for independent analysis
Each recording generates a feature set for exploratory longitudinal research.
Monitor change within a participant over time.
Repeated recordings provide an individual baseline. Subsequent measurements can be compared to detect statistically unusual deviations.
Example of the data provided for each recording (simulated for demonstration).
Each recording generates a feature set with optional within-subject baseline comparison.
| Metric | Current Value | Baseline | Z-score |
|---|---|---|---|
| Pitch Mean (F0) | 152.3 Hz | 148.7 Hz | +0.31 |
| Pitch Variability | 22.4 Hz | 19.8 Hz | +0.87 |
| Jitter | 0.031 | 0.028 | +0.52 |
| Shimmer | 0.082 | 0.075 | +0.44 |
| HNR | 18.3 dB | 20.1 dB | -0.65 |
| Temporal Variation | 0.021 | 0.018 | +0.29 |
| Spectral Centroid | 2147 Hz | 2098 Hz | +0.38 |
| Formant Variability Index | 412.6 | 389.2 | +0.71 |
| F0 Split-Half Consistency | 0.89 | 0.92 | -1.02 |
| Low/Mid Spectral Correlation | 0.67 | 0.71 | -0.53 |
⚠️ Baseline and Z-score values are shown when sufficient within-subject reference data are available. A ±2.58 threshold corresponds approximately to the central 99% range of a normal distribution and is provided for exploratory reference only.
All values shown are synthetic examples created for interface demonstration only. They do not represent measurements from an actual participant or validated clinical reference dataset.
Cogitrac Voice is designed for epilepsy research and longitudinal voice studies.
Explore quantitative changes in voice features in epilepsy research cohorts and longitudinal studies.
Track within-subject voice changes over time using an expanding baseline. Designed for observational studies.
Collect voice data and export metrics as CSV for independent analysis in R, Python, or other statistical software.
Cogitrac Voice is an investigational platform. Validation is in progress.
Voice samples are processed using Parselmouth/Praat and Librosa. Preprocessing includes RMS-based quality checks and pulse detection.
F0, jitter, shimmer, HNR, formants, spectral features, and 13 MFCCs are extracted and retained as candidate variables.
Full methodology →Individual baselines are established from repeated recordings. The number of recordings required should be determined by the study design and analysis plan.
Z-score-based deviation detection with a 99% reference range (±2.58 SD) where applicable. Values outside this range may be flagged for review.
Investigational research platform. Cogitrac Voice is not FDA-cleared, not clinically validated, and not intended for diagnosis, treatment, seizure prediction, or clinical decision-making. Individual features should not be interpreted as diagnostic, prognostic, treatment-related, or clinically actionable measurements. Always consult a qualified healthcare provider for any health-related decisions.
Researchers and participants retain control over their data.
Raw audio files are processed and numerical metrics are stored. Audio retention policies should be verified in the current platform documentation.
Export metrics as CSV for independent analysis. Raw audio is not included in standard exports.
Research data is de-identified and can be retained for ongoing studies. Participants can manage recordings through the platform.
Data is encrypted in transit and at rest. Access is restricted to authorized research personnel.
Privacy Policy →Create an account to explore the platform.
