5

Are low-pass filters and high-pass filters essentially the same when referring to accelerometer algorithms as when referring to sound engineering (audio processing)?

In sound engineering, a high pass filter cuts out the low frequencies associated with the bass sound, whereas low pass filters cut out high frequencies associated with treble sounds.

I want to understand what these filters are when applied to accelerometer data and how they are used, and am wondering if there's a parallel with the physics of sound. It's all physics, right?

If they are linked in some way that might allow me to understand how to measure accelerometer movements quicker than learning it from scratch.

Thanks

Dicky Moore
  • 956
  • 3
  • 10
  • 32
  • To answer your question: for a complementary filter approach, yes they are conceptually the same. But be aware that you may require very low cutoff frequencies for filtering motion data. Depending on your sampling rate the usual audio-rate (e.g. direct form biquad) filters do not perform well numerically at very low cutoff frequencies. You may find that some SVF algorithms (which are terrible with high-cutoff) are better. – Ross Bencina Sep 10 '13 at 14:09

1 Answers1

4

Yes, the concepts are exactly the same. If you think of frequency as how rapidly something changes, you will immediately see the parallels between filtering audio, images, sensor input - anything.

A low-pass filter will only allow relatively slow changes from its output. So any "jerkiness" in the accelerometer signal would be removed, and only the more gradual change (think overall trend) would pass.

Jonathon Reinhart
  • 132,704
  • 33
  • 254
  • 328
  • 1
    Thanks for your answer. So a high pass filter will give you output about things like shakes of the accelerometer and spikes like flicks, and a low pass will be for more slow movements, like scanning a landscape for augmented reality etc? – Dicky Moore Sep 08 '13 at 21:58
  • 1
    Yes, I think those are decent examples. See also [this related question](http://stackoverflow.com/questions/35530/what-are-high-pass-and-low-pass-filters). – Jonathon Reinhart Sep 09 '13 at 00:13
  • That's great, thank you. I think the thing I'm struggling with at the moment is how accelerometer data can be treated as a frequency. – Dicky Moore Sep 10 '13 at 13:14
  • 1
    Anything that changes w.r.t something else can be considered changing at some frequency. Your AC power changes from positive to negative polarity and back 60 times per second. `1/sec = 1 Hz` so we say the line frequency is 60 Hz. The (fundamental) pressure waves from the A key above middle C on a piano come and go 440 times per second, so they call it A = 440 Hz. – Jonathon Reinhart Sep 10 '13 at 22:48
  • 2
    Physical motion is no different. Now, because it's an accelerometer, things are a tad more complicated (as acceleration is the 2nd derivative of position), but let's pretend for a second that it's a position-ometer (like GPS) signal. If you moved your phone from left to right and back every second, the positional frequency of the phone would be 1 Hz. – Jonathon Reinhart Sep 10 '13 at 22:50