Citation
Jian Wu, Roozbeh Jafari. "Zero-Effort Camera-Assisted Calibration Techniques for Wearable Motion Sensors". International Conference on Cyber-Physical Systems (ACM/IEEE ICCPS), 14, April, 2014.

Abstract
Activity recognition using wearable motion sensors, also known as inertial measurement units (IMUs), plays an important role in wellness and healthcare monitoring applications. The design of the activity recognition algorithms typically requires information about the placement and the orientation of the IMUs on the body, and the signal processing is often designed, fitted and trained to work with a known orientation of the sensors. Signal processing techniques, and the placement of the sensors to capture physical movements are viewed as the cyber and physical systems, respectively. These two systems are required to remain tightly coupled to ensure the correct functionality of activity recognition algorithms, hence bringing a perfect perspective for cyber physical systems (CPS). In the case of accidental movement or displacement of the sensors, it is important to identify the new location and orientation to adjust the signal processing algorithms. This step, often called calibration or recalibration, requires extra effort on the part of the user to either perform a set of known movements, or enter information about the placement of the sensors manually. In this paper, we propose a camera-assisted calibration approach that does not require any extra efforts from the user. The calibration is done when the user is appearing in front of the camera and performing an arbitrary activity of choice (e.g., walking in front of the camera). Our approach constructs simulated IMU orientation readings, leveraging information acquired by a camera system. A two-step search algorithm is proposed to calibrate the sensor orientation by searching the minimum rotation distance between the simulated IMU and the real IMU readings. The novelty of our proposed methodology is in enabling seamless calibration techniques for wearable computers. Our experimental results show that the orientation of the sensors can be acquired with root mean square errors (RMSE) of 5.59 and 10.73 degrees for 6 daily movements when the sensor is placed on thigh and arm respectively.

Electronic downloads

• http://www.src.org/library/publication/p068850/
• http://dx.doi.org/10.1145/2668883.2668888

• HTML

  Jian Wu, Roozbeh Jafari. <a
  href="http://www.terraswarm.org/pubs/233.html"
  >Zero-Effort Camera-Assisted Calibration Techniques for
  Wearable Motion Sensors</a>, International Conference
  on Cyber-Physical Systems (ACM/IEEE ICCPS), 14, April, 2014.

• Plain text

  Jian Wu, Roozbeh Jafari. "Zero-Effort Camera-Assisted
  Calibration Techniques for Wearable Motion Sensors".
  International Conference on Cyber-Physical Systems (ACM/IEEE
  ICCPS), 14, April, 2014.

• BibTeX

  @inproceedings{WuJafari14_ZeroEffortCameraAssistedCalibrationTechniquesForWearable,
      author = {Jian Wu and Roozbeh Jafari},
      title = {Zero-Effort Camera-Assisted Calibration Techniques
                for Wearable Motion Sensors},
      booktitle = {International Conference on Cyber-Physical Systems
                (ACM/IEEE ICCPS)},
      day = {14},
      month = {April},
      year = {2014},
      abstract = {Activity recognition using wearable motion
                sensors, also known as inertial measurement units
                (IMUs), plays an important role in wellness and
                healthcare monitoring applications. The design of
                the activity recognition algorithms typically
                requires information about the placement and the
                orientation of the IMUs on the body, and the
                signal processing is often designed, fitted and
                trained to work with a known orientation of the
                sensors. Signal processing techniques, and the
                placement of the sensors to capture physical
                movements are viewed as the cyber and physical
                systems, respectively. These two systems are
                required to remain tightly coupled to ensure the
                correct functionality of activity recognition
                algorithms, hence bringing a perfect perspective
                for cyber physical systems (CPS). In the case of
                accidental movement or displacement of the
                sensors, it is important to identify the new
                location and orientation to adjust the signal
                processing algorithms. This step, often called
                calibration or recalibration, requires extra
                effort on the part of the user to either perform a
                set of known movements, or enter information about
                the placement of the sensors manually. In this
                paper, we propose a camera-assisted calibration
                approach that does not require any extra efforts
                from the user. The calibration is done when the
                user is appearing in front of the camera and
                performing an arbitrary activity of choice (e.g.,
                walking in front of the camera). Our approach
                constructs simulated IMU orientation readings,
                leveraging information acquired by a camera
                system. A two-step search algorithm is proposed to
                calibrate the sensor orientation by searching the
                minimum rotation distance between the simulated
                IMU and the real IMU readings. The novelty of our
                proposed methodology is in enabling seamless
                calibration techniques for wearable computers. Our
                experimental results show that the orientation of
                the sensors can be acquired with root mean square
                errors (RMSE) of 5.59 and 10.73 degrees for 6
                daily movements when the sensor is placed on thigh
                and arm respectively.},
      URL = {http://terraswarm.org/pubs/233.html}
  }
  

Posted by Barb Hoversten on 11 Dec 2013.