Using feature selection to evaluate pathological speech after training with a serious game
Loes van Bemmel, Catia Cucchiarini, Helmer Strik · ExLing Conferences · 2021
To evaluate the effectiveness of speech therapy, speech features before and after treatment can be compared, focussing on those features that changed most during treatment. In the current study acoustic features were automatically extracted from speech of patients affected by Parkinson’s Disease who had received speech treatment. Praat and openSMILE were used for feature extraction. Through feature selection, the top ten most characterizing features for pre vs. post-treatment were found. Further analysis of these features confirmed that after treatment the speakers spoke louder with lower pitch, which were the goals of the treatment.