# How to classify the movement type of a target into two categories(direct motion or else) by using a doppler radar?

How to classify the movement type of a target just into two categories(one is direct motion which is almost straight line movement and the other is anything else) by using a doppler radar with a microdoppler technic, or the spectrogram of the doppler frequency?

Could you recommend a paper, patent, and something like that related to this issue?

Thank you so much.

• You'd build a Doppler radar and estimate whether the Doppler shift is a trigonometric function of time. Standard Fourier analysis would be a solid baseline. – Marcus Müller Feb 21 '19 at 7:56
• It is very hard to do so if velocity of target is varying, with a single Doppler radar. With multiple it is possible. You can search for Doppler only tracking. – learner Feb 21 '19 at 9:42
• I just want to find whether the movement type of a target is approximately direct motion or not by using doppler spectrogram...Is there any paper, or something like that? – Sinecosine Feb 21 '19 at 11:51

If range isn’t important, sure you could do something like this with unmodulated pulses (i.e the pulsed you transmit are just sinusoids carriers with some known frequency). You would transmit the pulse and then receive it at some time delay. You could attempt to construct some sort of spectrogram or other time-frequency representation of the received signal. If the target is moving at a constant velocity, there would be a constant Doppler shift. If the target had non-constant velocity, you would be able to observe a non-constant Doppler shift.

If you cared about resolving range to the target, you would have to use a pulsed Doppler system. That would be very difficult to do with a pulsed Doppler system. Pulsed Doppler radars tend to ideally assume the target signal is moving at a constant rate, and that over the entire coherent processing interval (CPI), it doesn’t change range bins. The reason the assumption is key is because if the target shifts range bins, you will not get all of your Doppler processing gain and you’ll tend to smear out the target on the range-Doppler map, which makes detection of the target difficult if not impossible

In industry, the issue of range cell migration over a CPI for high velocity targets is pretty difficult but doable under the right circumstances. You can check out something called the keystone algorithm, that has seen some use in practice. I’d say it’d be very important to select your bandwidth and sample rate such that you don’t cause any extra apparent range migration of the target.

As for microdoppler on pulse Doppler radars , I have yet to actually see that be used in the industry I work in (defense), and that community largely treats it as computationally inefficient. With that being said, let’s say you have all the computation time in the world and you could do such a thing, sure it may reveal some characteristics. One classic example would be being able to detect the type of engine on some sort of airplane based on the microdoppler characteristics, e.g. the rotors of a helicopter or the supersonic wake of a jet.

In summary: yes, if you didn’t need range information you could analyze Doppler through using a spectrogram. Keep in mind that your RF carrier frequency will determine the Doppler shift that is apparent, so selecting an appropriate RF is pretty key for something like this.

• Thank you for your answer, actually it is difficult to understand your answer because of insufficient my knowledge,,, I thought that it is very simple task whether a target moves straight or not. For example, the doppler spectrogram of the target will be clear if the target moves toward or away from a radar as a direct motion with +10km/hrs velocity. The approximate straight line corresponding to the positive speed could be found in the doppler spectrogram of the target. OR, if the target does not move straight, we can't see the approximate straight line in the spectrogram. Maybe sinusoid... – Sinecosine Feb 22 '19 at 1:15
• @Sinecosine For a pulsed Doppler system that’s not really how it works. In your original question, can you add some system parameters? Do you care about range resolution at all? Are you only concerned with velocity? Are you using a Doppler radar or a PULSED Doppler radar? What are the target speeds? What RF are you propagating at? There are a lot of parameters that we’ll all need to be able to help you more – matthewjpollard Feb 22 '19 at 1:21
• Range resolution, velocity, RF type, type of doppler radar are not important. I just want to find whether a target moves straight or not by using doppler spectrogram. Thank you for your interest! – Sinecosine Feb 22 '19 at 1:45
• And target type are human, animal, vehicle, and something like that. – Sinecosine Feb 22 '19 at 1:47
• @Sinecosine thanks that helped clear up your question a bit, I’ve edited my answer. In short, yes with a non-pulsed Doppler radar, ie just a Doppler radar where you aren’t trying to determine range, you could do something like this. – matthewjpollard Feb 22 '19 at 1:56