On the use of magnetic field disturbances as features for activity recognition with on body sensors

We investigate the use of magnetic field disturbances as features for motion based, wearable activity recognition systems. Such disturbances are mostly caused by large metallic objects and electrical appliances, both of which are often involved in human activities. We propose to detect them by subtracting angular velocity values computed from the changes in the magnetic field vector from gyroscope signals. We argue that for activities that are associated with specific objects or devices such features increase system robustness against motion variations, sensor displacement and inter user differences. On a previously published data set of 8 gym exercises we demonstrate that our approach can improve the recognition by up to 31% over gyroscope only and up to 17% over a combination of a gyroscope and 3D accelerometer. Improvements of 9.5% are also demonstrated for user independent training as well as for the case of displaced sensors. A particularly interesting result is the fact that adding the magnetic disturbance features significantly improves recognition based on the vector norm of accelerometers and gyroscopes. The norm is often used when the orientation of the sensor is not known. This is common when using a mobile phone or other consumer appliance as a sensor.