Performance Evaluation of RNN with Hyperbolic Secant in Gate Structure through Application of Parkinson’s Disease Detection

This paper studies a novel recurrent neural network (RNN) with hyperbolic secant (sech) in the gate for specific medical application task of Parkinson’s disease (PD) detection. In detail, it focuses on fact that patients PD have motor speech disorders, by converting voice data into black-and-white images recurrence plot (RP) at time intervals and constructing detection model combines RNN convolutional (CNN); study evaluates performance sech compared long short-term memory (LSTM) gated unit (GRU) conventional gates. As result, proposed obtained similar results to LSTM GRU (an average accuracy about 70%) less hyperparameters, resulting faster learning. addition, framework gate, using tanh as output activation function is higher than relu function. The method will see more improvement increasing future. More analysis input sound type, RP image size, deep learning structures be included our future work further improving from voice.