![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8234844%2F62e83fe14b9268a635a23e00fdf418e7%2FScreenshot%20from%202022-07-09%2012-57-59.png?generation=1657340220439019&alt=media\")
I am looking into imu data.
\nLooking at the acceleration of the same TRIP smartphones.
\nGooglePixel4 and GooglePixel4XL appear to be oriented differently.
\nIn the graph, X is the acceleration in the E-axis (East) and Y is the acceleration in the N-axis (North) of the ENU coordinates.
\nGooglePixel4 appears to be moving in the southeast direction in accordance with TRIP.
\nGooglePixel4XL appears to be moving in the northwest direction.
\nIt appears that GooglePixel4XL is placed in the opposite direction to GooglePixel4.
\nPerhaps my calculations are wrong, but the process for the two smartphones is the same.
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8234844%2F673c60316251516d6340e0b1bb529cee%2FScreenshot%20from%202022-07-09%2015-26-43.png?generation=1657348189134318&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8234844%2Fa91bb57ef38483b0abf863876d2586db%2FScreenshot%20from%202022-07-09%2015-26-48.png?generation=1657348175051460&alt=media\")
I am looking into imu data.
\nLooking at the acceleration of the same TRIP smartphones.
\nGooglePixel4 and GooglePixel4XL appear to be oriented differently.
\nIn the graph, X is the acceleration in the E-axis (East) and Y is the acceleration in the N-axis (North) of the ENU coordinates.
\nGooglePixel4 appears to be moving in the southeast direction in accordance with TRIP.
\nGooglePixel4XL appears to be moving in the northwest direction.
\nIt appears that GooglePixel4XL is placed in the opposite direction to GooglePixel4.
\nPerhaps my calculations are wrong, but the process for the two smartphones is the same.
How did you get the X,Y,Z ENU accelerations from the X,Y,Z phone accelerations?
", "rawMarkdown": "How did you get the X,Y,Z ENU accelerations from the X,Y,Z phone accelerations?", "votes": null }, { "id": "1849961", "postDate": "07/10/2022 00:50:16", "content": "Thanks for the reply.
\nI used \"Orientation Filter + Stop Prediction with IMU\".
\nhttps://www.kaggle.com/code/queyrusi/orientation-filter-stop-prediction-with-imu
From this I get the quaternion and then the rotation matrix. Using the rotation matrix, I convert from IMU coordinates to NUE coordinates and calculate the acceleration. We multiplied the acceleration by the UTM time difference to obtain the velocity (which is only a relative velocity).
\nI am not sure if the calculation method is correct, but it is strange that the velocity of GooglePixel4 and GooglePixel4XL are opposite.
Yes, that uses the Mahony filter. I looked at it a bit - It seemed that while the orientation seemed reasonable most of the time, it sometimes was wildly inaccurate. It is possible that some other difference between the 2 phones is causing what you see. Have you tried looking at the raw X,Y,Z acceleration from the phones and comparing them? If they are from the same trip, shouldn't they line up in time and space correctly? Then you could see if they were opposite or not.
", "rawMarkdown": "Yes, that uses the Mahony filter. I looked at it a bit - It seemed that while the orientation seemed reasonable most of the time, it sometimes was wildly inaccurate. It is possible that some other difference between the 2 phones is causing what you see. Have you tried looking at the raw X,Y,Z acceleration from the phones and comparing them? If they are from the same trip, shouldn't they line up in time and space correctly? Then you could see if they were opposite or not.", "votes": null }, { "id": "1853041", "postDate": "07/12/2022 14:48:06", "content": "\n\nIf they are from the same trip, shouldn't they line up in time and space correctly?
\n
That's right. Thanks for the essential advice.
\nI didn't give up and made some adjustments and now it looks so similar (although there are still differences).
\nWhat I did was put in the smoothing process the raw X,Y,Z acceleration before putting Mahony filter.
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8234844%2F3f58a793d7f76d008a756de0397ba45b%2FScreenshot%20from%202022-07-12%2023-41-10.png?generation=1657637276998168&alt=media\")
I think the orientation of the smart phone, but the quality for the imu of different phones are not same and it maybe causes the error.
", "rawMarkdown": "I think the orientation of the smart phone, but the quality for the imu of different phones are not same and it maybe causes the error.", "votes": null }, { "id": "1861908", "postDate": "07/19/2022 10:31:59", "content": "I agree with you, I think it's because of the difference in imu for each phone and my poor processing.
", "rawMarkdown": "I agree with you, I think it's because of the difference in imu for each phone and my poor processing.", "votes": null } ], "comments": [ { "id": 1849827, "author_name": "solverworld", "author_url": "", "post_date": "07/09/2022 20:38:10", "content": "How did you get the X,Y,Z ENU accelerations from the X,Y,Z phone accelerations?
", "votes": null, "replies": [ { "id": 1849961, "author_name": "minfuka", "author_url": "", "post_date": "07/10/2022 00:50:16", "content": "Thanks for the reply.
\nI used \"Orientation Filter + Stop Prediction with IMU\".
\nhttps://www.kaggle.com/code/queyrusi/orientation-filter-stop-prediction-with-imu
From this I get the quaternion and then the rotation matrix. Using the rotation matrix, I convert from IMU coordinates to NUE coordinates and calculate the acceleration. We multiplied the acceleration by the UTM time difference to obtain the velocity (which is only a relative velocity).
\nI am not sure if the calculation method is correct, but it is strange that the velocity of GooglePixel4 and GooglePixel4XL are opposite.
Yes, that uses the Mahony filter. I looked at it a bit - It seemed that while the orientation seemed reasonable most of the time, it sometimes was wildly inaccurate. It is possible that some other difference between the 2 phones is causing what you see. Have you tried looking at the raw X,Y,Z acceleration from the phones and comparing them? If they are from the same trip, shouldn't they line up in time and space correctly? Then you could see if they were opposite or not.
", "votes": null, "replies": [] }, { "id": 1853041, "author_name": "minfuka", "author_url": "", "post_date": "07/12/2022 14:48:06", "content": "\n\nIf they are from the same trip, shouldn't they line up in time and space correctly?
\n
That's right. Thanks for the essential advice.
\nI didn't give up and made some adjustments and now it looks so similar (although there are still differences).
\nWhat I did was put in the smoothing process the raw X,Y,Z acceleration before putting Mahony filter.
I think the orientation of the smart phone, but the quality for the imu of different phones are not same and it maybe causes the error.
", "votes": null, "replies": [ { "id": 1861908, "author_name": "minfuka", "author_url": "", "post_date": "07/19/2022 10:31:59", "content": "I agree with you, I think it's because of the difference in imu for each phone and my poor processing.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1848953": "I am looking into imu data.\nLooking at the acceleration of the same TRIP smartphones.\nGooglePixel4 and GooglePixel4XL appear to be oriented differently.\nIn the graph, X is the acceleration in the E-axis (East) and Y is the acceleration in the N-axis (North) of the ENU coordinates.\nGooglePixel4 appears to be moving in the southeast direction in accordance with TRIP.\nGooglePixel4XL appears to be moving in the northwest direction.\nIt appears that GooglePixel4XL is placed in the opposite direction to GooglePixel4.\nPerhaps my calculations are wrong, but the process for the two smartphones is the same.\n\n\n\n\n\n", "1849827": "How did you get the X,Y,Z ENU accelerations from the X,Y,Z phone accelerations?", "1849961": "Thanks for the reply.\n\nI used \"Orientation Filter + Stop Prediction with IMU\". \nhttps://www.kaggle.com/code/queyrusi/orientation-filter-stop-prediction-with-imu\n\nFrom this I get the quaternion and then the rotation matrix. Using the rotation matrix, I convert from IMU coordinates to NUE coordinates and calculate the acceleration. We multiplied the acceleration by the UTM time difference to obtain the velocity (which is only a relative velocity).\nI am not sure if the calculation method is correct, but it is strange that the velocity of GooglePixel4 and GooglePixel4XL are opposite.", "1850668": "Yes, that uses the Mahony filter. I looked at it a bit - It seemed that while the orientation seemed reasonable most of the time, it sometimes was wildly inaccurate. It is possible that some other difference between the 2 phones is causing what you see. Have you tried looking at the raw X,Y,Z acceleration from the phones and comparing them? If they are from the same trip, shouldn't they line up in time and space correctly? Then you could see if they were opposite or not.", "1853041": ">If they are from the same trip, shouldn't they line up in time and space correctly? \n\nThat's right. Thanks for the essential advice.\nI didn't give up and made some adjustments and now it looks so similar (although there are still differences).\nWhat I did was put in the smoothing process the raw X,Y,Z acceleration before putting Mahony filter.\n\n", "1856751": "I think the orientation of the smart phone, but the quality for the imu of different phones are not same and it maybe causes the error.", "1861908": "I agree with you, I think it's because of the difference in imu for each phone and my poor processing." }, "source": "meta" }