Figure 1.

Exemplar overview of the work flow for the determination of the kinematics hyperparameters. First, the dataset consisting of 3D acceleration and 3D angular rate of five inertial sensors, and bilateral hip, knee and ankle joint 3D joint angles is split into training and test sets. Different preprocessing steps are then undertaken to create appropriate input and output data shapes for the training and testing of the three neural networks: MLP, LSTM and CNN.