I've just started a project which will involve having to detect certain events in a stream of kinematic sensor data. By searching through the literature, I've found a lot of highly specific papers, but no general reviews.

If I search up on computer vision, I'm likely to get 100s of articles giving overviews of different types of architectures for various vision tasks. They would look something like this:

  • We mainly use CNNs which work like this ...
  • For object detection we use one or two stage detectors which look like this...
  • For video classification we can use 3D CNNs or RNNs...
  • .... etc

So I'm looking for something similar with regard to kinematic motion sensors. As was pointed out to me on the signal processing SE, "kinematic" could mean a lot of things. So specifically, I'm referring to 1d time series data for:

  • acceleration/velocity/position
  • angular velocity / absolute orientation
  • 1
    $\begingroup$ 1D CNNs could help for time series. LSTM can be used to predict next value in a time series and anomaly detection. More info on patterns recognition in time series here - stackoverflow.com/questions/11752727/… $\endgroup$ – Stepan Novikov Jan 27 at 14:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.