Analýza řečových parametrů pro detekci čelistních anomálií

Abstract

This bachelor’s thesis focuses on the analysis of speech parameters for the detection of jaw anomalies, particularly sleep apnea. The theoretical section describes the essence of human speech, the current state of the issue, and various jaw anomalies. It also details the extracted speech parameters and the process of preprocessing the speech recording. Lastly, it introduces classification methods that can be used for the automatic classification of sleep apnea from speech. In the practical section, an algorithm for preprocessing and extracting speech parameters is designed in the MATLAB environment. Subsequently, classification methods are applied to these data using the MATLAB Classification Learner. A feature selection of extracted parameters is then implemented using the MRMR method. The effectiveness of this method is evaluated based on the accuracy and reduced computational intensity of the classification models as the number of analyzed parameters is progressively reduced. This work demonstrates the usefulness of analyzing speech parameters for detecting sleep apnea. It also validates feature selection methods in the context of reducing the computational intensity of classification models.