Human movement

[vitorsotero]

Hi,

I'm actually development a system that can evaluate human movements. I have a LSM9DS0 and a MPU6050 on hands.

I need to know not only wich movement was made, but the direction too. For this reason I think the LSM9DS0 is better (magnetometer).

I was reading about quaternion and rotation matrix, but it's a very complex subject for me. I just need the acceleration (x, y and z) calibrated. Any suggestion?

Thanks,

Vitor Sotero

[kriswiner]

If you just need acceleration then why the LSM9DS0?

If you need an absolute orientation estimation, then this is what you want:

https://www.tindie.com/products/onehorse/ultimate-sensor-fusion-solution/?pt=full_prod_search

On Sun, Sep 10, 2017 at 10:06 AM, vitorsotero notifications@github.com wrote:

Hi,

I'm actually development a system that can evaluate human movements. I have a LSM9DS0 and a MPU6050 in hands.

I need to know not only wich movement was made, but the direction too. For this reason I think the LSM9DS0 is better (magnetometer).

I was reading about quaternion and rotation matrix, but it's a very complex subject for me. I just need the acceleration (x, y and z) calibrated. Any suggestion?

Thanks,

Vitor Sotero

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qo66Pe8lseh8iC3dPDOmDG4HSlijks5shBcRgaJpZM4PSYaA .

[vitorsotero]

I need the velocity (I am integrating The acceleration). But the data fronteira a 9DOF sensor is more accurate... or not?

My big question is: How to use the fusion data to get the "calibrated acceleration"?

Em 10 de set de 2017 2:20 PM, "Kris Winer" notifications@github.com escreveu:

If you just need acceleration then why the LSM9DS0?

If you need an absolute orientation estimation, then this is what you want:

https://www.tindie.com/products/onehorse/ultimate- sensor-fusion-solution/?pt=full_prod_search

On Sun, Sep 10, 2017 at 10:06 AM, vitorsotero notifications@github.com wrote:

Hi,

I'm actually development a system that can evaluate human movements. I have a LSM9DS0 and a MPU6050 in hands.

I need to know not only wich movement was made, but the direction too. For this reason I think the LSM9DS0 is better (magnetometer).

I was reading about quaternion and rotation matrix, but it's a very complex subject for me. I just need the acceleration (x, y and z) calibrated. Any suggestion?

Thanks,

Vitor Sotero

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13, or mute the thread https://github.com/notifications/unsubscribe-auth/ AGY1qo66Pe8lseh8iC3dPDOmDG4HSlijks5shBcRgaJpZM4PSYaA .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-328357530, or mute the thread https://github.com/notifications/unsubscribe-auth/AeWh-Ww59FnRHrphIfQnNHFyieWL8CYxks5shBpugaJpZM4PSYaA .

[kriswiner]

What do you mean by calibrated acceleration? Do you mean linear acceleration, with gravity removed?

You need the quaternions for that, but it is straightforward.

On Sun, Sep 10, 2017 at 1:38 PM vitorsotero notifications@github.com wrote:

I need the velocity (I am integrating The acceleration). But the data fronteira a 9DOF sensor is more accurate... or not?

My big question is: How to use the fusion data to get the "calibrated acceleration"?

Em 10 de set de 2017 2:20 PM, "Kris Winer" notifications@github.com escreveu:

If you just need acceleration then why the LSM9DS0?

If you need an absolute orientation estimation, then this is what you want:

https://www.tindie.com/products/onehorse/ultimate- sensor-fusion-solution/?pt=full_prod_search

On Sun, Sep 10, 2017 at 10:06 AM, vitorsotero notifications@github.com wrote:

Hi,

I'm actually development a system that can evaluate human movements. I have a LSM9DS0 and a MPU6050 in hands.

I need to know not only wich movement was made, but the direction too. For this reason I think the LSM9DS0 is better (magnetometer).

I was reading about quaternion and rotation matrix, but it's a very complex subject for me. I just need the acceleration (x, y and z) calibrated. Any suggestion?

Thanks,

Vitor Sotero

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13, or mute the thread https://github.com/notifications/unsubscribe-auth/ AGY1qo66Pe8lseh8iC3dPDOmDG4HSlijks5shBcRgaJpZM4PSYaA .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-328357530, or mute the thread < https://github.com/notifications/unsubscribe-auth/AeWh-Ww59FnRHrphIfQnNHFyieWL8CYxks5shBpugaJpZM4PSYaA

.

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-328370248, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qkXv-AIw56IUQd6wQpyh1rSHEfUJks5shEjJgaJpZM4PSYaA .

[vitorsotero]

Linear acceleration and the direction

Em 10 de set de 2017 5:53 PM, "Kris Winer" notifications@github.com escreveu:

What do you mean by calibrated acceleration? Do you mean linear acceleration, with gravity removed?

You need the quaternions for that, but it is straightforward.

On Sun, Sep 10, 2017 at 1:38 PM vitorsotero notifications@github.com wrote:

I need the velocity (I am integrating The acceleration). But the data fronteira a 9DOF sensor is more accurate... or not?

My big question is: How to use the fusion data to get the "calibrated acceleration"?

Em 10 de set de 2017 2:20 PM, "Kris Winer" notifications@github.com escreveu:

If you just need acceleration then why the LSM9DS0?

If you need an absolute orientation estimation, then this is what you want:

https://www.tindie.com/products/onehorse/ultimate- sensor-fusion-solution/?pt=full_prod_search

On Sun, Sep 10, 2017 at 10:06 AM, vitorsotero notifications@github.com wrote:

Hi,

I'm actually development a system that can evaluate human movements. I have a LSM9DS0 and a MPU6050 in hands.

I need to know not only wich movement was made, but the direction too. For this reason I think the LSM9DS0 is better (magnetometer).

I was reading about quaternion and rotation matrix, but it's a very complex subject for me. I just need the acceleration (x, y and z) calibrated. Any suggestion?

Thanks,

Vitor Sotero

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13, or mute the thread https://github.com/notifications/unsubscribe-auth/ AGY1qo66Pe8lseh8iC3dPDOmDG4HSlijks5shBcRgaJpZM4PSYaA .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-328357530, or mute the thread < https://github.com/notifications/unsubscribe-auth/AeWh- Ww59FnRHrphIfQnNHFyieWL8CYxks5shBpugaJpZM4PSYaA

.

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-328370248, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qkXv- AIw56IUQd6wQpyh1rSHEfUJks5shEjJgaJpZM4PSYaA .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-328371194, or mute the thread https://github.com/notifications/unsubscribe-auth/AeWh-fVkOK-qSxg_dWzlgNH9fs-s_9JQks5shExbgaJpZM4PSYaA .

[vitorsotero]

Hi Kris!

From now I'm getting the quaternion from the LSM9DS0. I calibrated the magnetometer removing hard and soft Iron effects.

From now, I need to rotate my acceleration data (x, y and z). Do you think it's better to do with a rotation matrix (from quaternion) or the quaternion itself (qaccelDataq-¹)? Do you have any previous code?

My code to calibrate the mag:

void MagCalib() { while (calibrated != true)//calibrate magnetometer http://www.camelsoftware.com/2016/03/13/imu-maths-calculate-orientation-pt3/ { for (int i=0;i<10000;i++) { dof.readMag(); min_x=min(min_x,dof.mx); min_y=min(min_y,dof.my); min_z=min(min_z,dof.mz); max_x=max(max_x,dof.mx); max_y=max(max_y,dof.my); max_z=max(max_z,dof.mz); // Serial.println(i); // Serial.print(min_x);Serial.print(", ");Serial.print(min_y);Serial.print(", ");Serial.println(min_z); // Serial.print(max_x);Serial.print(", ");Serial.print(max_y);Serial.print(", ");Serial.println(max_z); }

mag_bias_x=(max_x + min_x)/2;//for hard iron error
mag_bias_y=(max_y + min_y)/2;
mag_bias_z=(max_z + min_z)/2;

max_scale_x = max_x - mag_bias_x; min_scale_x = min_x - mag_bias_x;
max_scale_y = max_y - mag_bias_y; min_scale_y = min_y - mag_bias_y;
max_scale_z = max_z - mag_bias_z; min_scale_z = min_z - mag_bias_z;

avg_x = max_scale_x + (min_scale_x*-1); avg_x = avg_x/2;
avg_y = max_scale_y + (min_scale_y*-1); avg_y = avg_y/2;
avg_z = max_scale_z + (min_scale_z*-1); avg_z = avg_z/2;
avg_total = avg_x + avg_y + avg_z; avg_total = avg_total/3;

scale_x = avg_total/avg_x; scale_y = avg_total/avg_y; scale_z = avg_total/avg_z;

//Serial.print("Hard error(x,y,z): ");Serial.print(mag_bias_x);Serial.print(", ");Serial.print(mag_bias_y);Serial.print(", ");Serial.println(mag_bias_z);
//Serial.print("Soft error(x,y,z): ");Serial.print(scale_x);Serial.print(", ");Serial.print(scale_y);Serial.print(", ");Serial.println(scale_z);
calibrated = true;

}

if (digitalRead(INT2XM)) { // When new magnetometer data is ready dof.readMag(); // Read raw magnetometer data // mx = dof.calcMag(dof.mx); // Convert to Gauss and correct for calibration // my = dof.calcMag(dof.my); // mz = dof.calcMag(dof.mz); mx = dof.mx;
my = dof.my; mz = dof.mz; mx -= mag_bias_x; //hard iron bias my -= mag_bias_y; mz -= mag_bias_z; mx = scale_x; //soft iron bias my = scale_y; mz *= scale_z; }

Thanks!!

[kriswiner]

Not sure I understand what you mean by "rotate my acceleration data".

On Sat, Sep 30, 2017 at 5:33 PM, vitorsotero notifications@github.com wrote:

Hi Kris!

From now I'm getting the quaternion from the LSM9DS0. I calibrated the magnetometer removing hard and soft Iron effects.

From now, I need to rotate my acceleration data (x, y and z). Do you think it's better to do with a rotation matrix (from quaternion) or the quaternion itself (qaccelDataq-¹)? Do you have any previous code?

My code to calibrate the mag:

void MagCalib() { while (calibrated != true)//calibrate magnetometer http://www.camelsoftware.com/2016/03/13/imu-maths- calculate-orientation-pt3/ { for (int i=0;i<10000;i++) { dof.readMag(); min_x=min(min_x,dof.mx); min_y=min(min_y,dof.my); min_z=min(min_z,dof.mz); max_x=max(max_x,dof.mx); max_y=max(max_y,dof.my); max_z=max(max_z,dof.mz); // Serial.println(i); // Serial.print(min_x);Serial.print(", ");Serial.print(min_y);Serial.print(", ");Serial.println(min_z); // Serial.print(max_x);Serial.print(", ");Serial.print(max_y);Serial.print(", ");Serial.println(max_z); }

mag_bias_x=(max_x + min_x)/2;//for hard iron error mag_bias_y=(max_y + min_y)/2; mag_bias_z=(max_z + min_z)/2;

max_scale_x = max_x - mag_bias_x; min_scale_x = min_x - mag_bias_x; max_scale_y = max_y - mag_bias_y; min_scale_y = min_y - mag_bias_y; max_scale_z = max_z - mag_bias_z; min_scale_z = min_z - mag_bias_z;

avg_x = max_scale_x + (min_scale_x-1); avg_x = avg_x/2; avg_y = max_scale_y + (min_scale_y-1); avg_y = avg_y/2; avg_z = max_scale_z + (min_scale_z*-1); avg_z = avg_z/2; avg_total = avg_x + avg_y + avg_z; avg_total = avg_total/3;

scale_x = avg_total/avg_x; scale_y = avg_total/avg_y; scale_z = avg_total/avg_z;

//Serial.print("Hard error(x,y,z): ");Serial.print(mag_bias_x);Serial.print(", ");Serial.print(mag_bias_y);Serial.print(", ");Serial.println(mag_bias_z); //Serial.print("Soft error(x,y,z): ");Serial.print(scale_x);Serial.print(", ");Serial.print(scale_y);Serial.print(", ");Serial.println(scale_z); calibrated = true;

}

if (digitalRead(INT2XM)) { // When new magnetometer data is ready dof.readMag(); // Read raw magnetometer data // mx = dof.calcMag(dof.mx); // Convert to Gauss and correct for calibration // my = dof.calcMag(dof.my); // mz = dof.calcMag(dof.mz); mx = dof.mx; my = dof.my; mz = dof.mz; mx -= mag_bias_x; //hard iron bias my -= mag_bias_y; mz -= mag_bias_z; mx = scale_x; //soft iron bias my = scale_y; mz *= scale_z; }

Thanks!!

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-333345236, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qhuQKcM20J5CQfvhK7jTzywTMhNvks5snt3zgaJpZM4PSYaA .

[vitorsotero]

I mean "rotate" in function to the Earth's reference, and not in function to the sensor reference. To correct the accelerometer data when the person makes a "randon" movement with the arm.

Em 30 de set de 2017 9:41 PM, "Kris Winer" notifications@github.com escreveu:

Not sure I understand what you mean by "rotate my acceleration data".

On Sat, Sep 30, 2017 at 5:33 PM, vitorsotero notifications@github.com wrote:

Hi Kris!

From now I'm getting the quaternion from the LSM9DS0. I calibrated the magnetometer removing hard and soft Iron effects.

From now, I need to rotate my acceleration data (x, y and z). Do you think it's better to do with a rotation matrix (from quaternion) or the quaternion itself (qaccelDataq-¹)? Do you have any previous code?

My code to calibrate the mag:

void MagCalib() { while (calibrated != true)//calibrate magnetometer http://www.camelsoftware.com/2016/03/13/imu-maths- calculate-orientation-pt3/ { for (int i=0;i<10000;i++) { dof.readMag(); min_x=min(min_x,dof.mx); min_y=min(min_y,dof.my); min_z=min(min_z,dof.mz ); max_x=max(max_x,dof.mx); max_y=max(max_y,dof.my); max_z=max(max_z,dof.mz ); // Serial.println(i); // Serial.print(min_x);Serial.print(", ");Serial.print(min_y);Serial. print(", ");Serial.println(min_z); // Serial.print(max_x);Serial.print(", ");Serial.print(max_y);Serial. print(", ");Serial.println(max_z); }

mag_bias_x=(max_x + min_x)/2;//for hard iron error mag_bias_y=(max_y + min_y)/2; mag_bias_z=(max_z + min_z)/2;

max_scale_x = max_x - mag_bias_x; min_scale_x = min_x - mag_bias_x; max_scale_y = max_y - mag_bias_y; min_scale_y = min_y - mag_bias_y; max_scale_z = max_z - mag_bias_z; min_scale_z = min_z - mag_bias_z;

avg_x = max_scale_x + (min_scale_x-1); avg_x = avg_x/2; avg_y = max_scale_y + (min_scale_y-1); avg_y = avg_y/2; avg_z = max_scale_z + (min_scale_z*-1); avg_z = avg_z/2; avg_total = avg_x + avg_y + avg_z; avg_total = avg_total/3;

scale_x = avg_total/avg_x; scale_y = avg_total/avg_y; scale_z = avg_total/avg_z;

//Serial.print("Hard error(x,y,z): ");Serial.print(mag_bias_x);Serial.print(", ");Serial.print(mag_bias_y);Serial.print(", ");Serial.println(mag_bias_z); //Serial.print("Soft error(x,y,z): ");Serial.print(scale_x);Serial.print(", ");Serial.print(scale_y);Serial.print(", ");Serial.println(scale_z); calibrated = true;

}

if (digitalRead(INT2XM)) { // When new magnetometer data is ready dof.readMag(); // Read raw magnetometer data // mx = dof.calcMag(dof.mx); // Convert to Gauss and correct for calibration // my = dof.calcMag(dof.my); // mz = dof.calcMag(dof.mz); mx = dof.mx; my = dof.my; mz = dof.mz; mx -= mag_bias_x; //hard iron bias my -= mag_bias_y; mz -= mag_bias_z; mx = scale_x; //soft iron bias my = scale_y; mz *= scale_z; }

Thanks!!

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-333345236, or mute the thread https://github.com/notifications/unsubscribe-auth/ AGY1qhuQKcM20J5CQfvhK7jTzywTMhNvks5snt3zgaJpZM4PSYaA .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-333345489, or mute the thread https://github.com/notifications/unsubscribe-auth/AeWh-Q3JQAHkHsiZoT7XvuFUOtuDmtJmks5snt_KgaJpZM4PSYaA .

[kriswiner]

Usually, with a 9 DoF sensor, the quaternions output from a proper fusion filter are absolute with respect to True North and Gravity. So I would derive anu accelerations you need from the quaternions.

On Sun, Oct 1, 2017 at 10:17 AM, vitorsotero notifications@github.com wrote:

I mean "rotate" in function to the Earth's reference, and not in function to the sensor reference. To correct the accelerometer data when the person makes a "randon" movement with the arm.

Em 30 de set de 2017 9:41 PM, "Kris Winer" notifications@github.com escreveu:

Not sure I understand what you mean by "rotate my acceleration data".

On Sat, Sep 30, 2017 at 5:33 PM, vitorsotero notifications@github.com wrote:

Hi Kris!

From now I'm getting the quaternion from the LSM9DS0. I calibrated the magnetometer removing hard and soft Iron effects.

From now, I need to rotate my acceleration data (x, y and z). Do you think it's better to do with a rotation matrix (from quaternion) or the quaternion itself (qaccelDataq-¹)? Do you have any previous code?

My code to calibrate the mag:

void MagCalib() { while (calibrated != true)//calibrate magnetometer http://www.camelsoftware.com/2016/03/13/imu-maths- calculate-orientation-pt3/ { for (int i=0;i<10000;i++) { dof.readMag(); min_x=min(min_x,dof.mx); min_y=min(min_y,dof.my); min_z=min(min_z, dof.mz ); max_x=max(max_x,dof.mx); max_y=max(max_y,dof.my); max_z=max(max_z, dof.mz ); // Serial.println(i); // Serial.print(min_x);Serial.print(", ");Serial.print(min_y);Serial. print(", ");Serial.println(min_z); // Serial.print(max_x);Serial.print(", ");Serial.print(max_y);Serial. print(", ");Serial.println(max_z); }

mag_bias_x=(max_x + min_x)/2;//for hard iron error mag_bias_y=(max_y + min_y)/2; mag_bias_z=(max_z + min_z)/2;

max_scale_x = max_x - mag_bias_x; min_scale_x = min_x - mag_bias_x; max_scale_y = max_y - mag_bias_y; min_scale_y = min_y - mag_bias_y; max_scale_z = max_z - mag_bias_z; min_scale_z = min_z - mag_bias_z;

avg_x = max_scale_x + (min_scale_x-1); avg_x = avg_x/2; avg_y = max_scale_y + (min_scale_y-1); avg_y = avg_y/2; avg_z = max_scale_z + (min_scale_z*-1); avg_z = avg_z/2; avg_total = avg_x + avg_y + avg_z; avg_total = avg_total/3;

scale_x = avg_total/avg_x; scale_y = avg_total/avg_y; scale_z = avg_total/avg_z;

//Serial.print("Hard error(x,y,z): ");Serial.print(mag_bias_x); Serial.print(", ");Serial.print(mag_bias_y);Serial.print(", ");Serial.println(mag_bias_z); //Serial.print("Soft error(x,y,z): ");Serial.print(scale_x); Serial.print(", ");Serial.print(scale_y);Serial.print(", ");Serial.println(scale_z); calibrated = true;

}

if (digitalRead(INT2XM)) { // When new magnetometer data is ready dof.readMag(); // Read raw magnetometer data // mx = dof.calcMag(dof.mx); // Convert to Gauss and correct for calibration // my = dof.calcMag(dof.my); // mz = dof.calcMag(dof.mz); mx = dof.mx; my = dof.my; mz = dof.mz; mx -= mag_bias_x; //hard iron bias my -= mag_bias_y; mz -= mag_bias_z; mx = scale_x; //soft iron bias my = scale_y; mz *= scale_z; }

Thanks!!

— You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-333345236 , or mute the thread https://github.com/notifications/unsubscribe-auth/ AGY1qhuQKcM20J5CQfvhK7jTzywTMhNvks5snt3zgaJpZM4PSYaA .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-333345489, or mute the thread https://github.com/notifications/unsubscribe-auth/AeWh- Q3JQAHkHsiZoT7XvuFUOtuDmtJmks5snt_KgaJpZM4PSYaA .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-333391721, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qtHO3kMR_qQ-hVVZywsYGrMUHPqpks5sn8kUgaJpZM4PSYaA .

[vitorsotero]

Hi, Kris!

I am now using quaternion to generate a rotation matrix. And then, I rotate the Accelerometer data to true North.

But that's a issue that I I'm facing. On the figure, there are 3 axes: xAccel_rotated (blue), yAccel_rotated (red) and zAccel_rotated (green). And two circles: the red was when I tilted the sensor on the W/E global system. The black when I tilted on the N/S global system.

Do you have an idea why this occurs (black circle)??

Here some code about my rotation matrix:

rotM[0] = 2.0f(q[0]q[0]) - 1 + 2.0f(q[1]q[1]); rotM[1] = 2.0f(q[1]q[2]) - 2.0f(q[3]q[0]); rotM[2] = 2.0f(q[1]q[3]) + 2.0f(q[2]q[0]); rotM[3] = 2.0f(q[1]q[2]) + 2.0f(q[3]q[0]); rotM[4] = 2.0f(q[0]q[0]) - 1 + 2.0f(q[2]q[2]); rotM[5] = 2.0f(q[2]q[3]) - 2.0f(q[1]q[0]); rotM[6] = 2.0f(q[1]q[3]) - 2.0f(q[2]q[0]); rotM[7] = 2.0f(q[2]q[3]) + 2.0f(q[1]q[0]); rotM[8] = 2.0f(q[0]q[0]) - 1 + 2.0f(q[3]q[3]); accel_R[1] = rotM[0]ax + rotM[1]ay + rotM[2]az; accel_R[2] = rotM[3]ax + rotM[4]ay + rotM[5]az; accel_R[3] = rotM[6]ax + rotM[7]ay + rotM[8]*az;

Thanks again,

Vitor Sotero

[kriswiner]

Hi Vitor,

I don't understand what you are doing...

On Wed, Oct 18, 2017 at 6:57 PM, vitorsotero notifications@github.com wrote:

Hi, Kris!

I am now using quaternion to generate a rotation matrix. And then, I rotate the Accelerometer data to true North.

But that's a issue that I I'm facing. On the figure, there are 3 axes: xAccel_rotated (blue), yAccel_rotated (red) and zAccel_rotated (green). And two circles: the red was when I tilted the sensor on the W/E global system. The black when I tilted on the N/S global system.

Do you have an idea why this occurs (black circle)??

Here some code about my rotation matrix: [image: f1] https://user-images.githubusercontent.com/31826425/31750635-041b3170-b460-11e7-9fb6-b0df647efb9f.PNG

rotM[0] = 2.0f(q[0]q[0]) - 1 + 2.0f(q[1]q[1]); rotM[1] = 2.0f(q[1]q[2]) - 2.0f(q[3]q[0]); rotM[2] = 2.0f(q[1]q[3]) + 2.0f(q[2]q[0]); rotM[3] = 2.0f(q[1]q[2]) + 2.0f(q[3]q[0]); rotM[4] = 2.0f(q[0]q[0]) - 1 + 2.0f(q[2]q[2]); rotM[5] = 2.0f(q[2]q[3]) - 2.0f(q[1]q[0]); rotM[6] = 2.0f(q[1]q[3]) - 2.0f(q[2]q[0]); rotM[7] = 2.0f(q[2]q[3]) + 2.0f(q[1]q[0]); rotM[8] = 2.0f(q[0]q[0]) - 1 + 2.0f(q[3]q[3]); accel_R[1] = rotM[0]ax + rotM[1]ay + rotM[2]az; accel_R[2] = rotM[3]ax + rotM[4]ay + rotM[5]az; accel_R[3] = rotM[6]*ax

• rotM[7]ay + rotM[8]az;

Thanks again,

Vitor Sotero

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-337777267, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qvefJNMGRJGmtC-N3bmzdDRxelhOks5stqyEgaJpZM4PSYaA .

[vitorsotero]

Summing up:

• I will evaluate human movement (using accelerometer data).
• I'm rotating the original accelerometer data using a rotation matrix from a quaternion.

[kriswiner]

I guess I will ask you the usual questions:

How did you calibrate your sensors?

Are you feeding the Madgwick function the data in the proper order?

I suspect even if the answers are yes, then somehow your rotation matrices are not quite right. But I am not sure i can check them for you...

On Wed, Oct 18, 2017 at 7:49 PM, vitorsotero notifications@github.com wrote:

Summing up:

• I will evaluate human movement (using accelerometer data).
• I'm rotating the original accelerometer data using a rotation matrix from a quaternion.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/LSM9DS0/issues/13#issuecomment-337784715, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1ql9FjZQBz0f8H0EUDXl3fvlN7KJwks5strjFgaJpZM4PSYaA .