I have a signal sampled at 100 sample per second. After low pass filtering of 200 order and some calculation such as zero crossing detection, I am getting around 1 second delayed signal. I want it in Realtime so that I have to predict signal 100 samples in future. I am using kalman filter to predict 2 samples ahead but I am confused wether I can use it or not to predict 100 samples in future. So that I can compensate delay introduced by low pass filter. What are the other methods available other than kalman filter.

  • 1
    $\begingroup$ I retract my answer. This is a one-to-one repost of your last question. $\endgroup$ – Marcus Müller Mar 15 '16 at 12:14
  • $\begingroup$ Which was rightfully closed as being unclear. You're not telling us anything about your signal! $\endgroup$ – Marcus Müller Mar 15 '16 at 12:15
  • $\begingroup$ I've read through your other questions on the stackexchange websites. Please start to learn from all the comments asking you to explicitly define what you're asking for and not leaving all the guesswork up to the reader. $\endgroup$ – Marcus Müller Mar 15 '16 at 12:18
  • $\begingroup$ I agree, you have to think about your questions for a longer time to make sure they pass the requirements of stackexchange. For example, you didnt say what kind of signal it was, if it is an engine, some music or a voice. you have to provide some context for us to think from, clear paragraphs, to state and develop your question as clearly as you can, it's good practice at writing. $\endgroup$ – DeltaEnfieldWaid Mar 15 '16 at 12:58

if you can determine the main sine frequencies that compose the lowpass signal throughput via fft or a filter bank, and their phase at the time of the signal (if last sample was rising or falling) you can predict as many samples ahead as their wavelength, for the bass component. as the signal rises and the prediction time is longer the prediction is worse obviously. So an FFT to see the frequencies of the bass components and an idea of their phase is ok to predict them roughly to the number of samples of their wavelength with certainty reducing over time. What is the overall context of the project?

  • $\begingroup$ 1 to 1 repost of last question. We shouldn't be answering, IMHO. $\endgroup$ – Marcus Müller Mar 15 '16 at 12:14

Not the answer you're looking for? Browse other questions tagged or ask your own question.