This article presents a wearable inertial measurement unit based system that estimates three-dimensional respiratory displacements on the thoracoabdominal surface. Such estimates can be useful in the calculation of respiratory rate, tidal volume, and for monitoring synchrony of compartments of the thoracoabdominal wall for diagnosis and physiotherapy. Challenges in estimating displacements from inertial sensors include drift due to sensor bias, errors due to the measurement of a large gravity component by accelerometers, and the need to compensate for real-time tilt angles from natural bending of the subject during breathing. An algorithm that combines double integration, high-pass filters, tilt angle estimation, and a Kalman smoother to estimate real-time gravity components along the sensor axes is utilized. The algorithm is facilitated by a special formulation of the gravity vector's kinematic model about the sensor axes. Preliminary experimental results show that the three-axes displacements are estimated with 1 mm accuracy at normal breathing frequencies of 0.25 Hz or higher, but lose some accuracy at lower breathing rates. The Kalman smoother which uses a combination of forward and backward filters is shown to provide significantly superior performance compared to a traditional Kalman filter, which only uses forward propagation. Additionally, the developed system also provides very accurate estimates of breathing frequency.
Bibliographical noteFunding Information:
This work was supported by the Pediatric Device Innovation Consortium, University of Minnesota
© 1996-2012 IEEE.
- Biomedical signal processing
- Kalman smoother
- inertial measurement units
- respiration monitoring
- wearable sensors