AVQI (Acoustic Voice Quality Index): A Clinician’s Guide to Scoring and Interpretation
Key takeaways
- AVQI is a single dysphonia-severity score, roughly 0 (normal) to 10 (severe), built from six acoustic measures.
- It combines a sustained vowel with connected speech, so it reflects everyday voice use better than vowel-only measures.
- Cut-offs depend on language: about 1.17 for American English and 2.35 for Italian in recent validation studies.
- Evidence is strong: a 2022 meta-analysis of 33 studies found pooled sensitivity of 0.83 and specificity of 0.89.
- Phonalyze reports AVQI automatically from live or remote recordings, with no app download for the patient.
What is the Acoustic Voice Quality Index (AVQI)?
The Acoustic Voice Quality Index (AVQI) is a multiparametric acoustic measure of overall dysphonia severity. It combines six acoustic parameters, taken from a sustained vowel and a sample of connected speech, into one score. Lower scores mean a healthier voice; higher scores mean more severe dysphonia.
AVQI was developed by Youri Maryn and colleagues in 2010 and is computed with a Praat script. The current and most widely validated version is AVQI 03.01 (sometimes called AVQI-3). It has since been validated in many languages, including English, Dutch, German, Italian, Finnish, Turkish and Lithuanian.
Single measures such as shimmer or HNR each describe one part of the voice signal. AVQI weighs several of them together, which makes it a better single number for tracking how a voice changes over therapy.
Why AVQI matters for SLPs and laryngologists
- One number to track. It summarises overall voice quality, which makes progress easier to chart across sessions.
- Closer to real speech. Adding connected speech captures voice onsets, offsets and variation that a held vowel misses.
- Matches what listeners hear. AVQI is designed to track perceptual dysphonia ratings such as the G (grade) in GRBAS and overall severity on CAPE-V.
- Objective documentation. It gives payers, referring physicians and patients a reproducible measure alongside perceptual judgement and patient-reported tools like the Voice Handicap Index.
The six acoustic components of AVQI 03.01
AVQI 03.01 draws on three kinds of analysis: cepstral, time-domain and frequency-domain.
| Measure | Domain | What it captures |
|---|---|---|
| CPPS (smoothed cepstral peak prominence) | Cepstral | How periodic and well-organised the voice is. Lower CPPS ↓ = more dysphonic. Learn more in CPP vs CPPS. |
| HNR (harmonics-to-noise ratio) | Time | Balance of harmonic energy against noise. Lower HNR ↓ = noisier, often breathier voice. |
| Shimmer local (%) | Time | Cycle-to-cycle amplitude variation, as a percentage. Higher ↑ = more instability. |
| Shimmer local (dB) | Time | The same amplitude variation, in decibels. Higher ↑ = more instability. |
| Slope of the long-term average spectrum | Frequency | How energy is spread between low and high frequencies. |
| Tilt of the regression line through the LTAS | Frequency | The overall spectral trend, which shifts with breathiness and glottal closure. |
The AVQI 03.01 formula
The six measures are combined with fixed weights, then rescaled so most voices fall between 0 and 10:
− 0.177 × CPPS
− 0.006 × HNR
− 0.037 × Shimmer local (%)
+ 0.941 × Shimmer local (dB)
+ 0.010 × Slope
+ 0.093 × Tilt ) × 2.8902
CPPS carries the largest weight among the noise-sensitive measures, which is why CPPS and AVQI usually move together. You don’t need to calculate it by hand: Phonalyze runs the analysis and puts AVQI in the voice report automatically.
How to interpret AVQI scores and cut-offs
A score above the validated cut-off suggests dysphonia. A rising score over time suggests worsening voice quality; a falling score suggests improvement.
| Language | Cut-off | Accuracy reported | Source |
|---|---|---|---|
| American English | 1.17 | AUC 0.84 · sensitivity 62% · specificity 95% | Castillo-Allendes et al., 2023 |
| Italian | 2.35 | — | Fantini et al., 2023 |
| All languages (meta-analysis) | Varies | Pooled sensitivity 0.83 · specificity 0.89 | Batthyany et al., 2022 |
Clinical insight: Cut-offs differ by language because the connected-speech sample differs. English validation uses Rainbow Passage text; Italian uses CAPE-V sentences. A score near the cut-off is a prompt for closer evaluation, not a diagnosis on its own.
Recording protocol for reliable AVQI
AVQI is sensitive to how the sample is recorded. These steps follow published validation protocols:
- Sustained vowel: the patient holds /a/ at comfortable pitch and loudness. The analysis uses a stable 3-second segment.
- Connected speech: the patient reads a standard text, such as the opening of the Rainbow Passage (English) or CAPE-V sentences. The speech segment is trimmed to roughly match the 3-second vowel.
- Quiet environment: background noise lowers CPPS and HNR and raises AVQI.
- Consistent microphone distance: keep it the same at every session so differences reflect the voice, not the setup.
- Same protocol every time: compare baseline and follow-up scores only when tasks, text and device are matched.
For more on how recording conditions affect results, see Audio & Voice Analysis with Phonalyze.
How to measure AVQI remotely with Phonalyze
Phonalyze is a browser-based voice analysis platform for speech-language pathologists and laryngologists. It runs Praat-based analysis in the cloud and includes AVQI in its voice reports, next to CPP/CPPS, harmonicity, jitter, shimmer, pitch, voicing and intensity.
- Create the session. Select the patient and add a sustained-vowel task and a connected-speech task, such as CAPE-V sentences.
- Send the anonymous link. For remote sessions, Phonalyze texts the patient an anonymous link. For live sessions, just start recording.
- Coach the setup. Ask for a quiet room, steady microphone distance, and a comfortable /a/ held for at least 3 seconds.
- Review the report. Results arrive in seconds, with AVQI shown alongside the individual measures that drive it.
- Track over time. Repeat the same tasks at each visit and pair AVQI with perceptual surveys such as VHI-10 or V-RQOL.
Patients need no app download, and patient data is stored encrypted on a HIPAA-compliant server. For a broader walkthrough, read How Therapists and SLPs Use Phonalyze for Voice Assessment.
Get AVQI in seconds, from anywhere
Send a link, get a full acoustic voice report with AVQI, CPPS, HNR and more. No downloads for you or your patients.
Start free 30-day trial Try the demoLimitations to keep in mind
- Language-specific: a cut-off validated in one language may not transfer to another.
- Device and room effects: phones, headsets and noisy rooms shift acoustic measures. Remote scores are most useful as trends within the same patient.
- Not a diagnosis: AVQI rates severity. It cannot identify the cause of dysphonia; laryngeal imaging is still needed.
- Limited data in some groups: evidence is thinner for children and some voice types, so interpret those scores with extra care.
- Pair with perception: use AVQI alongside CAPE-V or GRBAS and patient-reported outcomes, never alone.
Frequently asked questions about AVQI
What is a normal AVQI score?
Lower is better, but there’s no universal normal value. The cut-off depends on language and AVQI version: about 1.17 for American English speakers (Castillo-Allendes et al., 2023) and 2.35 for Italian (Fantini et al., 2023). Use the value validated for your patient’s language.
What does AVQI measure?
Overall dysphonia severity. AVQI 03.01 combines CPPS, HNR, shimmer local (%), shimmer local (dB), spectral slope and spectral tilt from a sustained vowel plus connected speech.
How accurate is AVQI?
A 2022 meta-analysis of 33 studies (Batthyany et al., Journal of Voice) found pooled sensitivity of 0.83 and specificity of 0.89, and concluded AVQI is a valid objective measure of dysphonia severity.
Why does AVQI use both a vowel and connected speech?
Held vowels give a stable signal but don’t reflect everyday voice use. Connected speech adds the onsets, offsets and prosody listeners hear. Together they match perceptual ratings more closely.
Can AVQI be measured remotely?
Yes. Phonalyze reports AVQI from recordings patients make on their own phone or computer through an anonymous browser link. Keep the setup consistent and read remote scores as trend data alongside perceptual evaluation.
Is AVQI better than CPPS alone?
CPPS is a strong single measure and a major part of AVQI. AVQI adds time- and frequency-domain measures that can make the score more robust across different dysphonia types. Many clinicians report both.
Related reading
References
- Maryn Y, Corthals P, Van Cauwenberge P, Roy N, De Bodt M. Toward improved ecological validity in the acoustic measurement of overall voice quality: combining continuous speech and sustained vowels. Journal of Voice. 2010;24(5):540–555.
- Maryn Y, De Bodt M, Roy N. The Acoustic Voice Quality Index: toward improved treatment outcomes assessment in voice disorders. Journal of Communication Disorders. 2010;43(3):161–174.
- Batthyany C, Barsties v. Latoszek B, Maryn Y. Meta-analysis on the validity of the Acoustic Voice Quality Index. Journal of Voice. 2022.
- Castillo-Allendes A, Codino J, Cantor-Cutiva LC, Nudelman CJ, Rubin AD, Barsties v. Latoszek B, Hunter EJ. Clinical utility and validation of the Acoustic Voice Quality and Acoustic Breathiness Indexes for voice disorder assessment in English speakers. Journal of Clinical Medicine. 2023;12(24):7679. doi:10.3390/jcm12247679
- Fantini M, et al. Validation of the Acoustic Voice Quality Index version 03.01 in Italian. Journal of Voice. 2023.
This article is for educational purposes for licensed clinicians and does not replace clinical judgement or laryngeal examination. Questions? Contact support@phonalyze.com.
— Phonalyze Team