IPIN ‘15 - A Comparative Analysis of Attitude Estimation for Pedestrian Navigation with Smartphones

This page has been created to supplement the article A Comparative Analysis of Attitude Estimation for Pedestrian Navigation with Smartphones submitted to IPIN 15’ conference.

Slides: here

Contact: Thibaud Michel


In order to have a variety of different handheld motions datasets, we identify five typical motions for a smartphone, inspired from Renaudin’s paper:

Acquisition platform

Platform used for our tests is a Google Nexus 5. Internal sensors are really cheap and cost at most 3€:
- a 6-axis InvenSense MPU6515 accelerometer and gyroscope
- a 3-axis AKM AK8963 magnetometer

Records of sensors has been made thanks to a homemade log application. InvenSense sensor can monitor activity up to 200Hz while AKM one only to 60Hz. For the purpose of aligning timestamps of magnetic field data with data obtained from accelerometer and gyroscope, we used a linear extrapolation, in order to keep the focus on a real-time algorithm, interpolation is unallowable here. We choose to align data at 200Hz but similar results can be obtained until a sampling at 50Hz.


An Allan variance study has been made on a large dataset. Results are shown on matlab .fig files


For our tests we used:

IPIN-calib-accelero.mat and IPIN-calib-magneto.mat are datasets used for accelerometer and magnetometer calibration.

The relation between uncalibrated data and calibrated data is given by:
calibrated_data = (uncalibrated_data - vector_bias) * matrix_trans

Platform Handler

A smartphone handler with infrared markers has been created with a 3D printer for this study.

Smartphone handler for Motion Lab with homemade recording app

Reference platform

Tests have been made in a 10mx10m square motion lab. In this room, we observed that the magnetic field is almost homogeneous from a subplace to another (variations are less than 3μT, and with negligible variations over time.

Heatmap of Magnetic Field of the Motion Lab

Our room is equipped with 20 Oqus cameras connected to a server and Qualisys Tracker software. A 150Hz sampling is used.

Kinovis room


Motion datasets are composed of:




Back Pocket



All results have been reported here