HNR (voice clarity)

Harmonics-to-Noise Ratio (HNR) measures how much of the signal energy is periodic (harmonics of the fundamental) versus aperiodic (noise, turbulence, aspiration). Higher HNR means a cleaner, more periodic signal; lower HNR means more noise or roughness. The tile on the dashboard is labelled "HNR (Voice Clarity)" to match the everyday meaning.

Reported value

Vocal Analyzer reports the mean HNR in dB across voiced frames. Typical values for clean speech-range recordings fall roughly between 10 and 25 dB. Values below about 7 dB usually indicate either a noisy recording or a breathy/rough voice source.

Why no gender bands

Published sex differences for mean HNR in speech are small — roughly 0–3 dB in the studies the analyzer bands are drawn from (Goy et al. 2013) — and overlap heavily between groups. A band overlay on that sort of difference would imply a level of resolution the measurement does not actually have, so Vocal Analyzer reports HNR without a gender classification.

Interpreting low HNR

Low HNR can mean several different things, and the interpretation depends on the rest of the dashboard:

• Noisy recording. The SNR warning at the top of the dashboard fires for the same reason — room noise, mic hiss, or a noisy recording chain put aperiodic energy into the signal.
• Breathy voice source. A breathy phonation produces audible aspiration, which reads as aperiodic energy and lowers HNR without meaning the recording is bad.
• Rough or creaky voice. Irregular cycle-to-cycle periodicity reads as low HNR even when the recording is clean.

Use CPPS (on the Voice quality tile) and the SNR warning together to disambiguate.

Caveats

• Not a health metric. Low HNR is a symptom of many things — some benign, some not — and Vocal Analyzer is not a clinical tool. If you are concerned about persistent hoarseness or voice discomfort, see a speech-language pathologist.
• Short recordings are noisy. HNR estimates stabilise with more voiced frames. A 3-second recording with a handful of voiced segments can produce an HNR value several dB off what a longer recording from the same speaker and mic would give.

References

• Goy, H., Fernandes, D. N., Pichora-Fuller, M. K., & van Lieshout, P. (2013). Normative voice data for younger and older adults. Journal of Voice 27(5), 545–555.
• Boersma, P. (1993). Accurate short-term analysis of the fundamental frequency and the harmonics-to-noise ratio of a sampled sound. Proceedings of the Institute of Phonetic Sciences 17, 97–110.