Fig. 5
From: Ensemble deep model for continuous estimation of Unified Parkinson’s Disease Rating Scale III
The architectures of the proposed deep models to estimate UPDRS-III score. a Dual-channel LSTM network to estimate UPDRS III from hand-crafted features. b 1D CNN-LSTM network to estimated UPDRS III from raw signals. Each convolutional layer is followed by a ReLU activation layer. Convolutional Block-2 was repeated to increase the depth of the CNN network. c. 2D CNN-LSTM network to estimate UPDRS III from time–frequency representations. The spectrogram of each 1-min window is the input to the CNN network. d The overall architecture of the proposed ensemble model