Removing gravity from accelerometer

[Mooneyg3]

Hi,

Is there an easy way to remove the gravity readings from the accelerometer?

Thanks,

[Gondwana]

Perhaps not a trivially easy way...

Convert the acceleration readings to a vector (magnitude and angles). Subtract g from the magnitude of the vector. Convert the modified vector back to (x,y,z).

I find that the inclusion of gravity is helpful because it lets me use accelerometer readings as de facto orientation readings, when there is no separate orientation sensor.

[Mooneyg3]

Thanks,

My math is aweful, looks like I'm going to have to read up on it.

What I'm looking to get is the actual velocity of wrist movement. What I
think I need to do is remove gravity from the readings and then use the xyz
velocity to calculate the actual velocity. So I don't think including
gravity will help me.

[YannisParis]

for velocity i would go with a vector as said :

convert xyz readings to a distance using this formula : 

var movement_total = Math.round((Math.round(10000 * Math.sqrt(accelerometer.x * accelerometer.x + accelerometer.y * accelerometer.y + accelerometer.z * accelerometer.z)) / 100))

then do this at a time T1 then at a time T2, substract the first measure to the second will automatically remove the gravity from the equation. 

edit : physicists would correct, it's not a distance since it's acceleration that is measured, but it will work. 

[Mooneyg3]

Thanks,

I don't think I am following though. The equation seems to be calculating the actual acceleration. What do the 10000 and 100 represent?

If acceleration is a constant 5m/s/s(accelerometer readings of 3,4,0). Are you saying the calculation measured ove1 second should be 

5*1 - 5*0 = 5.

So the velocity is 5m/s?

[YannisParis]

so the x10000 / 100 is only there to round to 2 digits after 0 (with math.round function), not useful in your case.

For your case, let's say that acceleration is 5m/s^2 as you said, basically after one second, your device would have move 5 meters from the starting point so the calculation is right (5m/s). The difficult part is to get to distances when acceleration is null, which depends on initial velocity and direction. 

For accounting for direction, I would calculate the delta of each coordinate acceleration, so a deceleration would be taken into account.

[Mooneyg3]

The issue I have there then is that Fitbit includes gravity. So instead of acceleration readings of 3,4,0 the reading would be 3,4,9.8 which works out at an acceleration of 11m/s/s and a velocity of 11m/s according to the above.

[YannisParis]

yes true- so the only solution is to get orientation, in order to know where the g value should be corrected at t0 and t1. Then 

From what I know, the Ionic is the only one with orientation/gyroscope