A method for turbulent noise estimation in voiced signals.

Area of Science:

• Acoustic analysis of voice
• Laryngeal function assessment
• Biomedical signal processing

Background:

• Assessing laryngeal function is crucial for diagnosing voice disorders.
• Existing acoustic parameters like Harmonic-to-Noise Ratio (HNR) and Normalized Noise Energy (NNE) have limitations due to their sensitivity to signal non-stationarity.
• A need exists for more robust acoustic parameters for voice analysis.

Purpose of the Study:

• To introduce and define a novel acoustic parameter, the Turbulent Noise Index (TNI).
• To evaluate TNI's effectiveness as an indicator of laryngeal function.
• To compare TNI's diagnostic performance against HNR and NNE for differentiating normal and pathological voices.

Main Methods:

• The Turbulent Noise Index (TNI) is defined as 100(1 - Rmax), where Rmax is the mean maximum correlation coefficient between consecutive glottal cycles.
• A calculation method for TNI is described.
• Experiments were conducted using synthetic and natural voice signals to assess TNI's properties and compare it with HNR and NNE.

Main Results:

• TNI demonstrates near independence from frequency and amplitude modulation noise.
• TNI is less affected by slow changes in frequency and amplitude compared to HNR and NNE.
• The overlap between normal and pathological voices was 14.8% for TNI, significantly lower than 21.5% for HNR and 23.5% for NNE.

Conclusions:

• TNI is a robust acoustic parameter for voice analysis.
• TNI offers a significant advantage over HNR and NNE as a diagnostic parameter for laryngeal function.
• The reduced overlap suggests TNI's superior ability to discriminate between normal and pathological voices.