Machine learning approach for detecting Covid-19 from speech signal using Mel frequency magnitude coefficient

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• The COVID-19 pandemic necessitates rapid and accessible diagnostic tools.
• Intelligent healthcare technologies offer potential for efficient disease detection.
• Speech signal analysis combined with AI presents a noninvasive diagnostic avenue.

Purpose of the Study:

• To design and develop a noninvasive, low-cost, remote diagnostic system for COVID-19 using speech signals.
• To investigate the efficacy of Mel frequency magnitude coefficients (MFMC) and machine learning for COVID-19 detection from voice.
• To compare the performance of different machine learning classifiers in identifying COVID-19 based on speech features.

Main Methods:

• Extraction of Mel frequency magnitude coefficients (MFMC) from speech signals to capture higher-order spectral features.
• Application of machine learning classifiers, including Random Forests and K-nearest neighbor (KNN), for COVID-19 detection.
• Evaluation of system performance using spectral coefficients (12, 24, 30, 40) and metrics like Area Under the Curve (AUC).

Main Results:

• Mel frequency magnitude coefficients (MFMC) demonstrated improved frequency resolution and reduced noise, enhancing diagnostic accuracy.
• The K-nearest neighbor (KNN) classifier achieved a superior Area Under the Curve (AUC) score of 0.80 compared to other models.
• Machine learning approaches proved effective and less computationally intensive than deep learning for this diagnostic task.

Conclusions:

• Speech signal analysis using MFMC and machine learning offers a viable noninvasive, low-cost method for remote COVID-19 detection.
• The KNN algorithm shows significant potential for accurate COVID-19 diagnosis based on vocal biomarkers.
• This intelligent healthcare approach can contribute to faster and more efficient pandemic response strategies.