Numerical Integration of accelerometer signal

Hellow,

I did sit to stand test with single IMU on the pelvis.

I have a acceleration data from IMU, and the units are m/s^2. The sample rate was 100Hz.

I did filter (butterwarth 4th order)

fc=3; % cutoff frequency

[b,a] = butter(4,fc/(100));

filtAcc = filter(b,a,y_acceleration);

and than numerical integration by cumtrapz function and remove the drift by detrend function twice: first time to get the velocity (units are [m/s] ?) and second time to get the displacement (units are [m] ?).

vertical_velocity=cumtrapz(filtAcc);

vertical_velocity_detrend=detrend(vertical_velocity);

vertical_displacement=cumtrapz(vertical_velocity_detrend);

vertical_displacement_detrend=detrend(vertical_displacement);

In the reality the displacement of the pelvis is 0.5 [m] or 50 [cm] why I got the displacement are in range -2000 to 2000 [m]?

Thank,

[Mathieu NOE]

double integration of accel signals are always sensitive to offset, drift and other issues

the one major issue I see in your code is that your are doing the cumtrapz integration without taking care of the sampling rate Fs

use dt = 1/Fs to correct it this way

vertical_velocity=dt*cumtrapz(filtAcc);