How to implement a LowPass Filter?

Question

I'm doing some math on both gyroscope and accelerometer data combined and I'd like to low pass filter the resulting data. So could someone post some generic code for a Low Pass filter?

Answer 1

A 1st order IIR low-pass filter can be of the form:

output_value = rate * input_value + (1.0 - rate) * previous_output_value;

which is pretty much what's inside Apple's AccelerometerGraph example. You select the rate parameter depending on what frequency (very very roughly shakes per second) you want to roll-off or start to attenuate to get a smoother resulting output, and the sample rate of the input data.

Answer 2

A low pass filter is simply smoothing of the results to remove the high frequencies. The simplest low pass filter is a box filter which is done by averaging n samples together.

For averaging 2 samples together this is as simple as doing:

sample[n] (sample[n] + sample[n + 1]) / 2;

Answer 3

If Apple's AccelerometerGraph example is too complex for you to understand, I created a simpler accelerometer example for my class which you can download here. This implements a simple low-pass and high-pass filter for raw accelerometer values, then logs the results to the screen.

As hotpaw2 and Goz describe, this uses a very simple weighted rolling average for the filter calculation:

UIAccelerationValue lowPassFilteredXAcceleration = (currentXAcceleration * kLowPassFilteringFactor) + (previousLowPassFilteredXAcceleration * (1.0 - kLowPassFilteringFactor));