0
$\begingroup$

I am emitting ultrasound with a speaker and am trying to track moving objects with a microphone array by detecting doppler shifts. I have succeeded to detect movement. However, to detect the direction of the movement, I found that I need to cancel the sounds reflected from the stationary objects. The goal is similar to MTI filters in aircraft radars which used for clutter filtering.

Consider input signal 1 (Sig1 for short) caused by movements (i.e., contains doppler shift) and input signal 2 (Sig2 for short) reflected from stationary objects. I have first simply tried subtracting the frequency spectra of Sig1 with Sig2 but did not work well. Furthermore, I tried to simply filter the non-doppler shift related frequencies by setting them to zero. This removes reflected sounds to a certain degree but still contains reflected sounds sources. Is there a robust way I can make use of?

I am relatively new to this area. Thank you for your help in advance.

| improve this question | |
$\endgroup$
  • $\begingroup$ it's not necessary to remove them in order to detect the direction, also i need to know more details about how you're going to use your Doppler signal (you could display its spectrum or you could play it on a load-speaker and detect the pitch by listening), and also about technique (pulsed Doppler or continuous, central frequency, length of interval you're able to record the signal and if there is any demodulation) $\endgroup$ – Mohammad M Jun 8 '18 at 23:05
  • $\begingroup$ If I don't remove them I get location of the sound reflected from nearby stationary objects with the direction of the detected motion. The speaker and microphone array are in the same position. $\endgroup$ – Deep Mind Jun 9 '18 at 2:12
  • $\begingroup$ The speaker and microphone array are in the same position. I am currently using continuous inaudible sound. Simply eliminating sound that does not have doppler shift (i.e., clutter spectrum) works well when the target is moving fast as the doppler shift is clearly separated. However, for slowly moving objects, the doppler shift ranges over the clutter spectrum. In such case removing clutter spectrum also impacts the doppler shift spectrum. $\endgroup$ – Deep Mind Jun 9 '18 at 2:25
  • $\begingroup$ @MohammadMohammadi, initially I started with implementing a program to localize sound source and used GCC PHAT. I continued to use this approach even after changing to sonar (through inaudible sound) based movement detection and found that it works to some extent. However, it seems the current approach I am using is not suitable for detecting slowly moving objects. I have tried different windows and tried but did not work. Is there a correlation approach that works better in this scenario? $\endgroup$ – Deep Mind Jun 26 '18 at 1:12

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.