Voice processing for COVID-19 scanning and prognostic indicator

Area of Science:

• Medical Science
• Acoustic Phonetics
• Infectious Diseases

Background:

• The COVID-19 pandemic necessitates reliable, non-contact diagnostic methods.
• Vocal correlates offer a potential avenue for detecting respiratory infections like COVID-19.

Purpose of the Study:

• To investigate voice parameter changes in COVID-19 patients.
• To explore the potential of voice analysis for COVID-19 screening and prognosis.

Main Methods:

• Thirty-six Asian ethnic volunteers (16 infected, 20 controls) vocalized vowels /a/, /e/, /i/, /o/, /u/.
• Voice recordings of infected patients during illness and after recovery were compared to non-infected controls.
• Acoustic analysis included fundamental frequency (F0), formant frequencies (F1-F4), Harmonic-to-Noise Ratio (HNR), jitter, and shimmer.

Main Results:

• Male COVID-19 patients showed higher energy intensity for /o/, F1 for /o/, and F3 for /u/ compared to controls.
• Female patients exhibited higher jitter for /o/ and /a/, but decreased F2 for /u/ and mean F0 for /e/, /i/, /o/.
• Significant differences in HNR, jitter, shimmer, and formant frequencies were observed between infected and recovered states, and between infected patients and controls.

Conclusions:

• Voice analysis reveals distinct acoustic changes associated with COVID-19 infection.
• Vocal parameters like fundamental frequency, formants, HNR, jitter, and shimmer can serve as reliable correlates for COVID-19.
• Developing a mobile application for real-time voice analysis could facilitate early detection and management of COVID-19.