Enhanced Human Action Recognition Using Fusion of Skeletal Joint Dynamics and Structural Features

Figure 3

An example of neural network architecture for the KTH dataset with 12 input neurons, 3 hidden layers, and 6 output neurons.