Deep learning detection approach for speech impairment children in Parkinson's disease
Ihab Ahmed Najm, Omar Gheni Abdulateef, Ahmed Hussein Ali, Saadaldeen Rashid Ahmed, Mohsin. A. Ahmed, Sameer Algburi
Abstract
Long-term neurological conditions that affect patients on a large scale, for instance, Parkinsonian disease (PD) represent a global strain in the life of patients and healthcare. Early detection and monitoring are the gates of successful treatment, these are the best ways that can be used to fight this disease and to stop its progress. The paper will be on the usefulness of voice pitch of tone of voice [rather than directly Wallace clear sounding as a marker of PD disease. The proposed paper looks forward to implementing a deep learning model to classify a varied arrangement of PD symptoms into a single class. As the two public datasets available for the data, the model is built with 61 fusing categories which there are 22 voice measures and 39 Mel Frequency Cepstral Coefficients (MFCC). The CNN-LSTM architecture that achieves a high accuracy rate of 97% was used. The brought forth model thus presents a viewing point for remote PD detection and monitoring which could be transformative to the clinical practice.