Adaptive Feature Mode Decomposition in Respiratory Sound Analysis

Accurate respiratory sound classification is essential to automated disease diagnosis. This study proposes an enhanced feature extraction method by integrating Mel-Frequency Cepstral Coefficients (MFCC) with Adaptive Features Mode Decomposition (AFMD). AFMD improves MFCC features by reducing noise, suppressing fluctuations, and enhancing feature stability. Experimental results on the ICBHI 2017 respiratory sound dataset show that AFMD significantly lowers the spectral entropy and standard deviation of MFCC coefficients, indicating a more structured and robust feature representation. This approach improves classification accuracy for respiratory diseases such as COPD, asthma, and pneumonia from 0.8641 to 0.9076, demonstrating its effectiveness in respiratory sound analysis. The proposed method offers a promising solution for automated disease screening and telemedicine applications.