Example of nonlinear dynamical parameters for machine learning-based speech pathology classifier
Aleksey Kharitonov, В. А. Антонец, K. N. Aleshin, Kirill N. Gromov · Journal of Physics Conference Series · 2019
In light of muscle coordinative structures might be defined as time-invariant dynamical systems that underlie an action's form, we find out if the ability of dynamical system to grade the levels of complexity according to external conditions is a feature of speech without sound disorders. Results revealed that such metrics of speech signal as correlation dimension and sample entropy as indicators of complexity vary widely under environmental conditions (acoustic fatigue) for participants without speech sound disorders, unlike for participants, who admitted the ones (sibilant /s/ considered). Supposedly, mentioned parameters may be of interest to machine learning-based speech pathology classifiers.