Evaluation of Mel-Frequency Cepstrum for Wheeze Analysis.

Area of Science:

• Medical Informatics
• Signal Processing
• Respiratory Medicine

Background:

• Wheeze monitoring is crucial for managing Chronic Respiratory Diseases like asthma and COPD.
• Automatic wheeze detection shows promise, with Mel-Frequency Cepstral Coefficients (MFCC) demonstrating effective performance.
• The specific contribution of individual MFCC coefficients to wheeze detection remains unclear.

Purpose of the Study:

• To characterize and evaluate the effectiveness of individual MFCC coefficients in differentiating wheeze and normal respiratory sounds.
• To determine which MFCC coefficients are most discriminative for wheeze detection.
• To compare the efficacy of single MFCC coefficients against combinations for wheeze identification.

Main Methods:

• Analysis of Mel-Frequency Cepstral Coefficients (MFCC) from respiratory sounds.
• Evaluation of coefficient effectiveness using histogram dissimilarity and linear separability metrics.
• Comparative analysis of single MFCC coefficients versus various combinations.

Main Results:

• The MFCC-2 coefficient demonstrated significantly higher effectiveness in discriminating between wheezes and normal respiratory sounds.
• Individual coefficient analysis revealed varying degrees of discriminative power.
• MFCC-2 outperformed other individual coefficients and combinations at an 8000 Hz sampling rate.

Conclusions:

• MFCC-2 is the most effective individual coefficient for detecting wheezes in respiratory sound analysis.
• This finding can optimize automatic wheeze detection algorithms for better chronic respiratory disease management.
• Targeted use of MFCC-2 can enhance the accuracy and efficiency of respiratory sound analysis systems.