Gabriele Ligorio

Estimating orientation using magnetic and inertial sensors and different sensor fusion approaches: accuracy assessment in manual and locomotion tasks.

Magnetic and inertial measurement units are an emerging technology to obtain 3D orientation of body segments in human movement analysis. In this respect, sensor fusion is used to limit the drift errors resulting from the gyroscope data integration by exploiting accelerometer and magnetic aiding sensors. The present study aims at investigating the effectiveness of sensor fusion methods under different experimental conditions. Manual and locomotion tasks, differing in time duration, measurement volume, presence/absence of static phases, and out-of-plane movements, were performed by six subjects, and recorded by one unit located on the forearm or the lower trunk, respectively. Two sensor fusion methods, representative of the stochastic (Extended Kalman Filter) and complementary (Non-linear observer) filtering, were selected, and their accuracy was assessed in terms of attitude (pitch and roll angles) and heading (yaw angle) errors using stereophotogrammetric data as a reference. The sensor fusion approaches provided significantly more accurate results than gyroscope data integration. Accuracy improved mostly for heading and when the movement exhibited stationary phases, evenly distributed 3D rotations, it occurred in a small volume, and its duration was greater than approximately 20 s. These results were independent from the specific sensor fusion method used. Practice guidelines for improving the outcome accuracy are provided.

Extended Kalman Filter-Based Methods for Pose Estimation Using Visual, Inertial and Magnetic Sensors: Comparative Analysis and Performance Evaluation

In this paper measurements from a monocular vision system are fused with inertial/magnetic measurements from an Inertial Measurement Unit (IMU) rigidly connected to the camera. Two Extended Kalman filters (EKFs) were developed to estimate the pose of the IMU/camera sensor moving relative to a rigid scene (ego-motion), based on a set of fiducials. The two filters were identical as for the state equation and the measurement equations of the inertial/magnetic sensors. The DLT-based EKF exploited visual estimates of the ego-motion using a variant of the Direct Linear Transformation (DLT) method; the error-driven EKF exploited pseudo-measurements based on the projection errors from measured two-dimensional point features to the corresponding three-dimensional fiducials. The two filters were off-line analyzed in different experimental conditions and compared to a purely IMU-based EKF used for estimating the orientation of the IMU/camera sensor. The DLT-based EKF was more accurate than the error-driven EKF, less robust against loss of visual features, and equivalent in terms of computational complexity. Orientation root mean square errors (RMSEs) of 1° (1.5°), and position RMSEs of 3.5 mm (10 mm) were achieved in our experiments by the DLT-based EKF (error-driven EKF); by contrast, orientation RMSEs of 1.6° were achieved by the purely IMU-based EKF.

A Novel Kalman Filter for Human Motion Tracking With an Inertial-Based Dynamic Inclinometer

Goal: Design and development of a linear Kalman filter to create an inertial-based inclinometer targeted to dynamic conditions of motion. Methods: The estimation of the body attitude (i.e., the inclination with respect to the vertical) was treated as a source separation problem to discriminate the gravity and the body acceleration from the specific force measured by a triaxial accelerometer. The sensor fusion between triaxial gyroscope and triaxial accelerometer data was performed using a linear Kalman filter. Wrist-worn inertial measurement unit data from ten participants were acquired while performing two dynamic tasks: 60-s sequence of seven manual activities and 90 s of walking at natural speed. Stereophotogrammetric data were used as a reference. A statistical analysis was performed to assess the significance of the accuracy improvement over state-of-the-art approaches. Results: The proposed method achieved, on an average, a root mean square attitude error of 3.6° and 1.8° in manual activities and locomotion tasks (respectively). The statistical analysis showed that, when compared to few competing methods, the proposed method improved the attitude estimation accuracy. Conclusion: A novel Kalman filter for inertial-based attitude estimation was presented in this study. A significant accuracy improvement was achieved over state-of-the-art approaches, due to a filter design that better matched the basic optimality assumptions of Kalman filtering. Significance: Human motion tracking is the main application field of the proposed method. Accurately discriminating the two components present in the triaxial accelerometer signal is well suited for studying both the rotational and the linear body kinematics.

Dealing with Magnetic Disturbances in Human Motion Capture: A Survey of Techniques

Magnetic-Inertial Measurement Units (MIMUs) based on microelectromechanical (MEMS) technologies are widespread in contexts such as human motion tracking. Although they present several advantages (lightweight, size, cost), their orientation estimation accuracy might be poor. Indoor magnetic disturbances represent one of the limiting factors for their accuracy, and, therefore, a variety of work was done to characterize and compensate them. In this paper, the main compensation strategies included within Kalman-based orientation estimators are surveyed and classified according to which degrees of freedom are affected by the magnetic data and to the magnetic disturbance rejection methods implemented. By selecting a representative method from each category, four algorithms were obtained and compared in two different magnetic environments: (1) small workspace with an active magnetic source; (2) large workspace without active magnetic sources. A wrist-worn MIMU was used to acquire data from a healthy subject, whereas a stereophotogrammetric system was adopted to obtain ground-truth data. The results suggested that the model-based approaches represent the best compromise between the two testbeds. This is particularly true when the magnetic data are prevented to affect the estimation of the angles with respect to the vertical direction.

Prior-to- and Post-Impact Fall Detection Using Inertial and Barometric Altimeter Measurements

This paper investigates a fall detection system based on the integration of an inertial measurement unit with a barometric altimeter (BIMU). The vertical motion of the body part the BIMU was attached to was monitored on-line using a method that delivered drift-free estimates of the vertical velocity and estimates of the height change from the floor. The experimental study included activities of daily living of seven types and falls of five types, simulated by a cohort of 25 young healthy adults. The downward vertical velocity was thresholded at 1.38 m/s, yielding 80% sensitivity (SE), 100% specificity (SP) and a mean prior-to-impact time of 157 ms (range 40-300 ms). The soft falls, i.e., those with downward vertical velocity above 0.55 m/s and below 1.38 m/s were analyzed post-impact. Six fall detection methods, tuned to achieve 100% SE, were considered to include features of impact, change of posture and height, singularly or in association with one another. No single feature allowed for 100% SP. The detection accuracy marginally improved when the height change was considered in association with either the impact or the change of posture; the post-impact fall detection method that analyzed the impact and the change of posture together achieved 100% SP.

How Angular Velocity Features and Different Gyroscope Noise Types Interact and Determine Orientation Estimation Accuracy.

In human movement analysis, 3D body segment orientation can be obtained through the numerical integration of gyroscope signals. These signals, however, are affected by errors that, for the case of micro-electro-mechanical systems, are mainly due to: constant bias, scale factor, white noise, and bias instability. The aim of this study is to assess how the orientation estimation accuracy is affected by each of these disturbances, and whether it is influenced by the angular velocity magnitude and 3D distribution across the gyroscope axes. Reference angular velocity signals, either constant or representative of human walking, were corrupted with each of the four noise types within a simulation framework. The magnitude of the angular velocity affected the error in the orientation estimation due to each noise type, except for the white noise. Additionally, the error caused by the constant bias was also influenced by the angular velocity 3D distribution. As the orientation error depends not only on the noise itself but also on the signal it is applied to, different sensor placements could enhance or mitigate the error due to each disturbance, and special attention must be paid in providing and interpreting measures of accuracy for orientation estimation algorithms.

Assessing the Performance of Sensor Fusion Methods: Application to Magnetic-Inertial-Based Human Body Tracking.

Information from complementary and redundant sensors are often combined within sensor fusion algorithms to obtain a single accurate observation of the system at hand. However, measurements from each sensor are characterized by uncertainties. When multiple data are fused, it is often unclear how all these uncertainties interact and influence the overall performance of the sensor fusion algorithm. To address this issue, a benchmarking procedure is presented, where simulated and real data are combined in different scenarios in order to quantify how each sensor’s uncertainties influence the accuracy of the final result. The proposed procedure was applied to the estimation of the pelvis orientation using a waist-worn magnetic-inertial measurement unit. Ground-truth data were obtained from a stereophotogrammetric system and used to obtain simulated data. Two Kalman-based sensor fusion algorithms were submitted to the proposed benchmarking procedure. For the considered application, gyroscope uncertainties proved to be the main error source in orientation estimation accuracy for both tested algorithms. Moreover, although different performances were obtained using simulated data, these differences became negligible when real data were considered. The outcome of this evaluation may be useful both to improve the design of new sensor fusion methods and to drive the algorithm tuning process.

A novel functional calibration method for real-time elbow joint angles estimation with magnetic-inertial sensors

Magnetic-inertial measurement units (MIMUs) are often used to measure the joint angles between two body segments. To obtain anatomically meaningful joint angles, each MIMU must be computationally aligned (i.e., calibrated) with the anatomical rotation axes. In this paper, a novel four-step functional calibration method is presented for the elbow joint, which relies on a two-degrees-of-freedom elbow model. In each step, subjects are asked to perform a simple task involving either one-dimensional motions around some anatomical axes or a static posture. The proposed method was implemented on a fully portable wearable system, which, after calibration, was capable of estimating the elbow joint angles in real time. Fifteen subjects participated in a multi-session experiment that was designed to assess accuracy, repeatability and robustness of the proposed method. When compared against an optical motion capture system (OMCS), the proposed wearable system showed an accuracy of about 4° along each degree of freedom. The proposed calibration method was tested against different MIMU mountings, multiple repetitions and non-strict observance of the calibration protocol and proved to be robust against these factors. Compared to previous works, the proposed method does not require the wearer to maintain specific arm postures while performing the calibration motions, and therefore it is more robust and better suited for real-world applications.

A linear Kalman Filtering-based approach for 3D orientation estimation from Magnetic/Inertial sensors

The accurate estimation of the three dimensional (3D) orientation estimation from Magnetic/Inertial Measurement Units (MIMUs) is a challenging task due to the noisiness of the sensor data and the non-linearity of the measurement models. Recently, new linear Kalman Filtering-based (KF) estimators have been presented in literature which address the tilt angles estimation problem (i.e. the pitch and roll angles, or the attitude) as a source separation technique applied to the accelerometer signal. In this paper one of these methods is extended to the magnetometer signal, under the assumption of hard-iron magnetic errors. The Earths magnetic field is then estimated in a linear KF framework to provide an additional reference for heading estimation, yielding full 3D orientation estimation. The proposed method was validated on data from a body-worn MIMU. Five subjects and two scenarios were included in the experimental validation. The proposed KF lowered the magnetic errors to less than 4 μT, with corresponding orientation errors that ranged from 2.8° (attitude) to 8.5° (heading).

A Simulation Environment for Benchmarking Sensor Fusion-Based Pose Estimators.

In-depth analysis and performance evaluation of sensor fusion-based estimators may be critical when performed using real-world sensor data. For this reason, simulation is widely recognized as one of the most powerful tools for algorithm benchmarking. In this paper, we present a simulation framework suitable for assessing the performance of sensor fusion-based pose estimators. The systems used for implementing the framework were magnetic/inertial measurement units (MIMUs) and a camera, although the addition of further sensing modalities is straightforward. Typical nuisance factors were also included for each sensor. The proposed simulation environment was validated using real-life sensor data employed for motion tracking. The higher mismatch between real and simulated sensors was about 5% of the measured quantity (for the camera simulation), whereas a lower correlation was found for an axis of the gyroscope (0.90). In addition, a real benchmarking example of an extended Kalman filter for pose estimation from MIMU and camera data is presented.