SVM-based Efficient Sleep States Classification and Management Derived from Breath Sound Measurements

Breath sounds (BS) contain important physiological indicators, making their analysis and detection a well-established area of study. This paper proposes a method for identifying and classifying sleep stages based on the spectral power and spectral flux of breath sounds. These multiple features effectively capture the characteristics of breath sounds, enabling efficient classification of data related to both normal and abnormal breath sounds. We employ support vector machine (SVM) classification models to achieve automatic, high-efficiency classification using the spectral characteristics of breath sounds. The high classification accuracy obtained validates the performance of the proposed feature sets and classification model. Additionally, the experimental results demonstrate that breath sounds can be used to evaluate sleep states as an alternative to commercial products.