I just can't wrap my head around this so I decided to ask.
Imagine that you have a 3D array with 6 layers. Each of these layers contain data about 6 people demonstrating 6 different movements 10 times. Each movement varies in length, the longest lasting 800 time steps. The movements of the joints are saved as x,y,z coordinates. So, for 1 layer, the columns are x,y,z coordinates and time_steps. How would one feed this data to a recurrent neural net for movement classification? The input shape has to be (number of samples, number of time steps in each sample, number of variables in each sample).
Would one need to create a 3D array for this task to feed to the network in the shape (6x6x10, 800 - time steps for the duration of 1 movement, 3 - x,y,z coordinates)?
If you are familiar with time series classification using neural nets, how would you approach this task?
