Biomedical Engineering Reference
In-Depth Information
in the upper limb, and the Modified Ashworth Scale (MAS) (Ashworth 1964) for
measurement on muscle spasticity.
In our previous work on the poststroke wrist training assisted by the interac-
tive robot, several robotic parameters were used to monitor the recovery process
during a training program (Hu
et al
. 2008; Hu
et al
. 2009). They were:
(1)
EMG Activation Levels of Muscles
: EMG activation level mainly reflects
signal amplitude, and is related to the muscle power, or force, generated in a
task. Besides the EMG signals of ECR and FCR muscles, another two muscles
related to the elbow joint, i.e. the biceps brachii (BIC) and the triceps brachii
(TRI) muscles were also studied in the training program. The raw EMG signals
of these four muscles during wrist tracking tasks were recorded and stored in a
computer. Then, the linear envelope of the recorded EMG signals was obtained
by
(i) full-wave rectification,
(ii) low-pass filtering (10Hz cutoff frequency with forth-order, zero-phase
forward and reverse Butterworth filter),
(iii) subtraction of the average EMG activity during the resting state, and
(iv) normalized to the maximum value of EMG activation in each muscle
during either a training session or the IMVE and IMVE of each session.
The EMG activation level of a muscle in a tracking trial was calculated by
averaging the EMG envelope of the trial.
(2)
Muscle Cocontraction Indexes (CI) of Muscle Pairs
: After obtaining the
EMG envelope of the muscles of interest, we also calculated the coactivations
between different muscle pairs during the tracking task, which could be
expressed by the cocontraction index as introduced in our previous studies
(Hu
et al
. 2007; Hu
et al
. 2008; Hu
et al
. 2009), that is,
1
T
=
A
ij
(
)
CI
t
dt
(5.4)
T
where
A
ij
(t) was the overlapping activity of EMG linear envelopes for muscles
i
and
j
,and
T
was the length of the signal. The value of a CI for a muscle pair
varied from 0 (no overlapping at all in the signal trial) to 1 (total overlapping
of the 2 muscles with both EMG levels kept at 1 during the trial). The repre-
sentative segments of EMG envelopes obtained in the tracking experiment are
pattern in a motion task.
(3)
The root-mean-squared-error (RMSE) of tracking trajectories
: The root-
mean-squared-error of a trial calculates the difference between the target and
the actual elbow angles during the tracking task. This measure represents the
performance of a tracking task, i.e. the small the RMSE value, the better the
accuracy of performance. An example of the measured actual wrist angle, in
