User-Adaptive Inertial Sensor Network for Feedback-Controlled Gait Support Systems

Inertial sensor networks on the lower limbs enable realtime assessment of gait and thus feedback control of active gait support systems. However, most state-of-the-art methods (1) require each sensor unit to be attached to a predefined segment in a predefined orientation, or (2) require the user to perform precise calibration motions, and (3) require a homogeneous magnetic field. Such requirements are incompatible with most clinical applications especially if the patient attaches the sensor autonomously. We propose methods for a plug-and-play gait analysis that uses only accelerometer and gyroscope readings. These methods allow the sensor network to adjust to the user and to calibrate itself automatically using data from only five seconds of walking, i.e. the methods are plug-and-play. In particular, we present a sensor-to-segment pairing method that identifies which sensor is attached to which body segment (thigh, shank and foot) of which leg. Analysis of data from over 500 trials with healthy subjects and Parkinson’s patients yields a correct-pairing success rate of 99.8%. Additionally, we present a sensor-to-segment calibration method that determines the knee joint axis in the local coordinates of each sensor, which is essential for joint angle calculation. Comparing the resulting knee joint angles to measurements of an optical motion capture system yields root-mean-square deviations of about 3◦.