Acoustic Voice Quality Index (AVQI)

AVQI is an acoustic measure of voice quality used in clinical and research practices. It combines six acoustic markers from connected speech and sustained vowel vocal tasks to yield a single robust measure of dysphonia. Importantly, AVQI correlates well with auditory-perceptual judgements of voice quality, exhibits high test-retest reliability, and demonstrates high sensitivity to voice quality changes through voice therapy. Bonus – it is quick and easy to measure using the Praat freeware!


To measure AVQI, first download the AVQI text file (download here) and ensure that you have already installed Praat. Then complete the below three easy steps!

First, record the following sentence directly into Praat (or upload to Praat): “When light strikes raindrops in the air, they act like a prism and form a rainbow.” Save this audio file as “cs” (for connected speech). It is important you save it with this title.

Side note: emerging research suggests that smartphone microphones positioned ~10 cm from the mouth can capture acceptable recordings for acoustic signal analysis

Second, record into Praat (or upload to Praat) a sustained /a/ for five seconds. From this file, extract the middle 3 seconds, and rename it as “sv” (sustained vowel). It is important you save it with this title.

Lastly, select “Praat”, then “New Praat Script”, then copy and paste the AVQI text file (downloaded from the beginning of this tutorial) into your script window, then select “Run”, “Run”, and “OK”

And just like that, you have measured AVQI! In the above example, there was an AVQI of 1.00.


Like smoothed Cepstral Peak Prominence (and unlike jitter, shimmer, and harmonics-to-noise ratio), the AVQI is a valid and reliable acoustic voice measure. This makes it highly valuable for both clinical and research practices, when evaluating a patient and tracking changes over time.


A higher number correlates with a more dysphonic voice quality. Vocally healthy adults exhibit an average AVQI of ~2.3 (SD: 0.8; Min-Max: 0.6-5.0). Furthermore, it has been found that an AVQI of ≥3.31 exhibiting a sensitivity of 71% and a specificity of 88% for detecting a voice disordered patient. See references for more details and age- and sex-based norms, along with AVQI scores across voice disorders!


References:

  • Latoszek, B. B. V., Ulozaitė-Stanienė, N., Maryn, Y., Petrauskas, T., & Uloza, V. (2019). The influence of gender and age on the acoustic voice quality index and dysphonia severity index: a normative study. Journal of Voice, 33(3), 340-345.

  • Barsties v. Latoszek, B., Ulozaitė‐Stanienė, N., Petrauskas, T., Uloza, V., & Maryn, Y. (2019). Diagnostic accuracy of dysphonia classification of DSI and AVQI. The Laryngoscope, 129(3), 692-698.

  • Maryn, Y., & Weenink, D. (2015). Objective dysphonia measures in the program Praat: smoothed cepstral peak prominence and acoustic voice quality index. Journal of Voice, 29(1), 35-43.

  • Maryn, Y., De Bodt, M., Barsties, B., & Roy, N. (2014). The value of the Acoustic Voice Quality Index as a measure of dysphonia severity in subjects speaking different languages. European Archives of Oto-Rhino-Laryngology, 271(6), 1609-1619.

  • Faham, M., Laukkanen, A. M., Ikävalko, T., Rantala, L., Geneid, A., Holmqvist-Jämsén, S., ... & Pirilä, S. (2019). Acoustic voice quality index as a potential tool for voice screening. Journal of Voice.

  • Woerd, B., Wu, M., Vijay, P., Doyle, P.C., & Fung, K. (2020). Evaluation of acoustic analyses of voice in nonoptomized conditions. Journal of Speech, Language, and Hearing Research, 63(12), 3991-3999. https://doi.org/10.1044/2020_JSLHR-20-00212

  • Shuman, E., Awan, S., Rosen, C. A., Schneider, S. L. (2020, October 24). Remote Voice Testing: A comparison of cepstral analysis between two remote modalities versus in-person recording. Fall Voice Conference, University of California San Francisco.

Acknowledgments: I'd like to thank Adrián Castillo-Allendes for supplying me with the AVQI script and Sarah Schneider for references regarding remote voice testing. Thank you!