Skip to main content
. 2020 Aug 15;20(16):4581. doi: 10.3390/s20164581

Figure 1.

Figure 1

Scheme of the workflow. The marker trajectories of gait trials, saved as .c3d files, were used to calculated inertial sensor data and the joint angles of the lower limbs. The inertial data were divided into five subsets and used as input to train long short-term memory neural networks that were used to predict the joint angles. To adapt its weights and biases for the prediction, the network received the calculated joint angles as target values.