Fig. 2
From: Assessing inertial measurement unit locations for freezing of gait detection and patient preference
One-dimensional convolutional neural network model architecture. Two-layer, one-dimensional convolutional neural networks were trained using 2-s windows of raw inertial measurement unit (IMU) data. Each convolutional layer had 16 filters with a kernel length of 17. Weights were trained with 50% dropout. Convolutional layers were followed by a max pooling layer with a pool length of 2 and a 10-node fully-connected layer. The output of the model was the probability that the majority label for the 2-s window of raw IMU data was freezing of gait (FOG)