Fundamental frequency measures in the comparison of habitual and disguised voices.
Amanda de Castro Matheus, Denise de Oliveira Carneiro Berejuk, Eliane Cristina Pereira and 2 others
Base f0, minimum f0, and maximum f0 held steady when speakers disguised their voices, while mean and median f0 shifted upward.
For forensic voice comparison, the stable measures are the ones to rely on when a suspect has altered their pitch.
Key findings
1Base f0, minimum f0, and maximum f0 showed no statistically significant differences between habitual and disguised speech (p = 0.333, p = 0.751, p = 0.968), meaning these measures were not altered by the speakers' disguise strategies.
2Mean f0 (habitual 198.85 Hz vs disguised 216.62 Hz, p = 0.021) and median f0 (habitual 195.31 Hz vs disguised 206.1 Hz, p = 0.032) were significantly higher in disguised speech, with effect sizes of 0.38 and 0.35 (small to medium).
3The most common disguise strategies were strained voice and greater loudness (each used by 21 of 40 speakers, 52.5%), followed by lower pitch (18 of 40, 45%) and phoneme distortions, omissions, or substitutions (16 of 40, 40%).
Still to come
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study did not measure natural f0 variability in undisguised speech, so it cannot separate how much of the observed change is due to disguise versus normal fluctuation between two readings. All samples were read text in a controlled acoustic booth, not spontaneous or semi-spontaneous speech, so the findings may not generalise to the phone calls and environmental recordings that forensic work actually involves. The study tested whether measures are stable within one speaker; it did not measure accuracy, sensitivity, or specificity for telling two different speakers apart. The voice bank was collected for a previous study, and the disguise instruction was a single generic prompt ('try not to be recognized'), so the range of disguise strategies may be narrower than what appears in real criminal cases.
Declared interests
No funding source or conflict-of-interest declaration is stated in the supplied text.
The easy way to misread this
Do not read the stability of base f0 as proof it can identify a speaker. This study only shows that base f0 does not change when the same person disguises their voice. It does not test whether base f0 can tell two different speakers apart, and the authors note that accuracy, sensitivity, and specificity have not been measured for any of these parameters.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →