Parkinson Disease Prediction Using CNN-LSTM Model from Voice Signal

Parkinson disease (PD) is a neurodegenerative disease cause by the lack of dopamine hormone secretion. Humans get affected in motor and non-motor activities with Parkinsonism. Motor dysfunction affects speech production. Speech is produced by the muscles of the larynx, trachea, epiglottis, vocal fold, vocal tract, tongue, pallet, and cartilage. It has been studied that PD can be identified by looking at changes in speech signals over time. The deep learning based features have been used for effective characterization of speech signal. In this paper, a hybrid CNN-LSTM classifier is proposed to efficiently detect the PD. To assess the performance of the proposed approach, 22 healthy and 28 Parkinson patients who speak Italian are used. An average accuracy of 97% is achieved with the proposed method. The results suggest that the proposed approach is appropriate for automatic identification of PD in practical scenarios.