Glottal Airflow Estimation using Neck Surface Acceleration and Low-Order Kalman Smoothing.

Area of Science:

• Bioengineering
• Signal Processing
• Acoustic Analysis

Background:

• Non-invasive vocal function measurement using skin accelerometers over the extrathoracic trachea is an emerging technique.
• Current methods for estimating glottal airflow, such as inverse filtering or Bayesian techniques, rely on subglottal impedance models but suffer from sensor positioning errors and model mismatch, leading to computational burdens.
• Accurate glottal airflow estimation is crucial for understanding vocal function and diagnosing voice disorders.

Purpose of the Study:

• To develop a computationally efficient method for estimating glottal airflow using system identification and a Kalman smoother.
• To reduce the model order and computational requirements compared to existing Bayesian approaches.
• To enable real-time estimation of glottal airflow and its uncertainty for wearable monitoring applications.

Main Methods:

• System identification techniques were employed to obtain a low-order state-space representation of the subglottal impedance-based model.
• A Kalman smoother was utilized with the reduced-order model to estimate glottal airflow from neck-surface acceleration data.
• The proposed method was compared against previous Bayesian techniques in terms of accuracy, model order, and computational time.

Main Results:

• The proposed approach achieved a 94% reduction in model order and required only 1.5% of the computing time compared to prior Bayesian methods.
• The Kalman smoother demonstrated slightly improved accuracy in correcting for glottal airflow deviations.
• The method provides a valuable measure of uncertainty in the airflow estimate, adaptable to varying measurement conditions.

Conclusions:

• The developed Kalman smoother approach offers a significant reduction in computational load and model complexity for glottal airflow estimation.
• The method achieves comparable accuracy to existing techniques while providing uncertainty quantification.
• This approach holds promise for real-time, accurate, and uncertainty-aware glottal airflow estimation in wearable voice monitoring devices.