I know that here we just want to reimplement the Matlab version of the code. However, in the INDIP algorithm, the yaw angle is calculated from the pelvis orientation using a Madgwick filter, which improves the yaw estimation as gyro drift is corrected with the information from the accelerometer. Maybe, for future testing, we could implement a yaw_estimation_method argument to let the user decide if she or he wants to use the raw integration or the sensor fusion method.
I know that here we just want to reimplement the Matlab version of the code. However, in the INDIP algorithm, the yaw angle is calculated from the pelvis orientation using a Madgwick filter, which improves the yaw estimation as gyro drift is corrected with the information from the accelerometer. Maybe, for future testing, we could implement a yaw_estimation_method argument to let the user decide if she or he wants to use the raw integration or the sensor fusion method.
_Originally posted by @pltsc18 in https://github.com/mobilise-d/mobgap/pull/131#discussion_r1622553353_