The development of microelectromechanical system components in the last decade
enabled the production of small, inexpensive, and low-power sensors. These sensors
present a low-cost, wearable, and easier-to-use alternative to expensive laboratory
measurement systems which can be used in many different fields such as medicine,
biomechanics, sport, sociology, physiology and engineering. Typical representative of
wearable sensors are inertial and magnetic measurement units (IMUs). The raw signals
are processed using sensory fusion to assess the orientation of the sensor and
the corresponding segment, on which the sensor is mounted. One of most popular
methods of sensory fusion is the extended Kalman filter (EKF). The basic principle
of Kalman filtering in orientation estimation is based on obtaining an orientation estimate
by integrating angular velocity. The estimate is further fused with orientation
estimated from the measured gravitational acceleration (inclination) and magnetic field
(heading). Two major issues emerge by implementing this principle : i) the accelerometer
on the IMU measures the resulting difference between gravitational and dynamical
acceleration, and ii) the weak Earth’s magnetic field can easily become disturbed in
the vicinity of ferromagnetic materials and electromagnetic devices.
For estimating the orientation of individual segments of the human body based
on inertial and magnetic measurements, we developed an EKF-based sensory fusion
method. We implemented the model of the measured acceleration as a combination
of known translational acceleration of the mounting point, gravitational acceleration,
and acceleration caused by rotation. The latter includes the kinematic description of
the placement of the sensors on the segment. Similarly we can define the acceleration
of the end point of the segment, which is at the same time the acceleration of the
mounting point of the following segment. By implementing the recursive calculation
of the orientation, only the acceleration of the mounting point of the first segment in
the kinematic chain must be known.
The Earth’s magnetic field is used as a reference vector for angle estimation around
the vertical axis. To prevent the effect of a disturbed magnetic field on the orientation
estimation, we proposed active compensation of magnetic disturbances. The method is
based on estimating the magnetic disturbance by considering current measured magnetic
field, measured magnetic field at previous time step, and the change of orientation in
current time step. The measured magnetic field can be thus modeled as a combination
of the Earth’s magnetic field and the assessed magnetic disturbance.
We modeled the human body with a 7 segments model. Each segment was equipped
with an individual IMU. The recursive algorithm for the calculation of segment
orientations was implemented based on a serial kinematic chain. The foot, which was
in contact with the floor, was assumed to be the first segment of the chain with zero
acceleration. Real-time determination of the initial segment was based on the reactive
force data from the measurement insoles. The force assessed with the insoles was also
used as an input into the recursive Newton-Euler calculation of joint torques.
The performance of the algorithm for orientation estimation of single segment was
experimentally evaluated. A single pendulum was equipped with the wearable and a
reference sensory system. The movement analysis showed that the median absolute
error of the assessed angles was below 5◦.
Sit-to-stand (STS) transfer is a short-term maneuver. In cased where we are interested
in monitoring more complex movement (STS being just one of the many maneuvers)
long-term reliability of the algorithm must be ensured. The proposed system was
experimentally validated in a long-term walking on a treadmill and on a circular polygon
with stairs, simulating different activities in everyday life (level-ground walking,
stair negotiating, turning). Comparison of assessed and reference kinematic parameters
yielded a median absolute error of the assessed angles of 5◦, with no expressed drift
over time, regardless of the movement dynamics, duration, or type of the performed
maneuver.
The experimental validation of the performance of the magnetic compensation algorithm
was accomplished. The experimental setup comprised of a single pendulum,
mimicking one segment, and solenoidal coil which was used to introduce magnetic disturbances
into the system. The results showed a large deviation of the angles, assessed
without magnetic compensation (error up to 200◦). The orientations, assessed by algorithm
with active compensation of the magnetic disturbances, were more precise with
only a moderate error, regardless of the strength and direction of the imposed magnetic
disturbance.
Sit-to-stand transfer is one of the most common movements of daily life. It is assumed
to be symmetrical with respect to body sagittal plane for healthy individuals,
while persons with movement disabilities stand up in an asymmetrical way. Sit-tostand
analysis of subjects following transtibial amputation showed evident asymmetry
in kinematic and dynamic parameters with the sound limb being exposed to higher
stress. The influence of different seat heights and velocities on asymmetry was tested.
The asymmetry was not affected neither in kinematic nor dynamic parameters.
This indicates that asymmetry in standing up pattern of subjects following transtibial
amputation is the result of the standing up pattern developed after the amputation.
We used the wearable sensory system to perform an analysis of STS of the patient
with left-hip arthrosis. The patient was scheduled for a total hip replacement. Before
the surgical procedure, the subject’s STS pattern was prominently asymmetrical in
terms of joint angles and torques. Analysis of standing up three months and half
a year after the surgery showed improved STS transfer with substantial asymmetry
reduction between the affected and the sound side. This experiment confirmed the
potential of wearable sensors being used in clinical practice for monitoring the effect
of a surgical procedure and for assessing the success of the rehabilitation.
The wearable sensory system is composed of inertial and magnetic measurement
units, measurement insoles, and appropriate sensory fusion algorithms. It represents
an alternative solution for measuring kinematic and dynamic parameters of human
movement with moderate error (median error < 5◦). The error of the assessed quantities
is not affected by the measuring environment, dynamics, duration, or type of the
monitored movement. The system with implemented magnetic compensation can also
be used in the environment with disturbed magnetic field.
We present the solutions for two major issues of orientation estimation with an
IMU: i) drift of the assessed orientation during long-term measurement and ii) effect of
magnetic disturbances on the orientation assessment. By implementing the solutions
we enabled the use of the wearable sensory system as a precise measurement instrument
for measurement and analysis of human motion in clinical practice. In addition the
sensory system can be used for providing feedback to the user or as a part of control systems of wearable robots.
|
