Generalized Linear Model With the DTW distance

The significant indices from the 4 categories of DTW distance were used to distinguish between severity levels (mild or moderate I). These indices were then used as explanatory variables to create generalized linear models (GLMs) for each vowel. A 5-fold cross-validation method with 110 waveform samples was used to train the model for each vowel, which was then used to predict the severity classification. R was used for GLM modeling and label prediction. The pROC package (version 1.18.0) for R was used to obtain the receiver operating characteristic (ROC) curve and calculate the area under the curve (AUC), whereas confusion matrices were generated using the Caret package (version 6.0) for R.