Analysis of COVID-19 Resulting Cough Using Formants and Automatic Speech Recognition System

Area of Science:

• Biomedical Engineering
• Signal Processing
• Infectious Disease Research

Background:

• The COVID-19 pandemic necessitates novel diagnostic tools.
• Cough is a primary symptom of respiratory infections, including COVID-19.
• Acoustic analysis of coughs offers a non-invasive method for potential disease detection.

Purpose of the Study:

• To develop and validate a cough recognition system for differentiating COVID-19 infected individuals from healthy ones.
• To investigate the utility of acoustic features like formants and pitch in cough analysis for COVID-19 detection.
• To contribute to research on leveraging speech recognition technologies for pandemic response.

Main Methods:

• Implementation of a Hidden Markov Model (HMM) based cough recognition system.
• Utilizing Mel-Frequency Cepstral Coefficients (MFCC) as feature vectors (13 dimensions, 39 dimensions overall).
• Comparative analysis of formant frequencies (F0, F1, F2, F3, F4) and pitch between COVID-19 positive and healthy coughs.

Main Results:

• The HMM system achieved a 6.7% difference in recognition rates between infected and non-infected individuals.
• Significant variations in formant frequencies (F1, F3, F4) were observed in the coughs of COVID-19 infected individuals.
• Lower variations were noted for F0 and F2 formants, suggesting specific acoustic markers.

Conclusions:

• The developed HMM-based cough recognition system shows potential in distinguishing COVID-19 infected individuals.
• Formant frequency analysis provides valuable insights into acoustic changes associated with COVID-19 infection.
• Acoustic analysis of coughs can serve as a supplementary tool in the context of COVID-19 diagnosis and monitoring.