Automated System Using HMM for Lung Disease Recognition Based on Cough Sounds

• Published in: Computational engineering and physical modeling Vol. 8; no. 1; pp. 25 - 35
• Main Authors: Shing-Tai Pan, Bo-Kai Lee, You-Qian Wu
• Format: Journal Article
• Language: English
• Published: Pouyan Press 01.01.2025
• Subjects: hidden markov model (hmm); mel-frequency cepstral coefficients (mfcc) feature extraction; speech recognition
• ISSN: 2588-6959
• DOI: 10.22115/cepm.2024.490612.1350

The main purpose of this paper is to recognize multiple lung diseases simultaneously when a cough sound is detected. In addition to detecting whether the person producing the cough sound has a lung disease, it can also identify which disease the person has. This paper will use Hidden Markov Models (HMM) and Delta Mel-Frequency Cepstral Coefficients (MFCC) feature extraction to recognize the disease of a cough. Cough sound is a biological marker that can be used to detect diseases by observing the differences in waveforms between healthy and sick individuals. Previous research has used cough sounds to identify single lung diseases and mental illnesses. However, they cannot achieve high recognition rates and low computation and necessitate extensive training datasets. Therefore, this study aims to design a model that can recognize multiple lung diseases with low computation and high recognition rate. In this paper, a hyperparameter optimization procedure is performed. By this procedure, the results can have better performance and achieve a good recognition rate.