Are Speech Attractor Models Useful in Diagnosing Vocal Fold Pathologies?

Development of non-invasive methods in diagnosingdifferent diseases can lead to improvement of preventionand care programs. the speech is an easily accessiblesignal, which clearly represents the characteristics of larynxand vocal folds. therefore, application of some propermachine learning algorithms (e.g., feature extractionand classification methods) on a small part of a recordedspeech signal may help in diagnosing vocal fold diseasessuch as paralysis, edema, nodules, and polyp. generally, transforming the input signal into the set of features iscalled feature extraction. if these features are accuratelyextracted, it is expected that the feature set will capturethe relevant pathological information of speech signal to predict the diagnosis.