Detection of larger signals

I would like to know if it’s possible to detect signal larger than the size of my training window, in cases like accelerometers, if I have samples of lets say 1second and I want to be able to recognize a motion that happens on 2 seconds, would I be able to detect that or my neural network will run 2 different inferences?
And if not do I need to scale my window size to the size of that large signal? Because if so it would add some latence to the inference.
Thanks guys
Project ID:

Context/Use case:

Hi @JosuGaztelu,

The window size must be at least as large as the signal you are trying to detect. If your signal is, say, 2 seconds, it will not fit into the 1 second input for the impulse (DSP + NN). As a result, it will not be classified correctly.

There are a few ways to remedy this:

  • Increase the size of your window
  • Lower the sampling rate (if you don’t need to detect high frequency components) and increase the size of the window
  • Detect parts of the whole signal (for example, if a 1 second piece of your 2 second signal can be uniquely classified as belonging to that class, then you might not need all 2 seconds)

Hope that helps!