A total of 6 exercise conditions were
executed. The exercise protocol was
block randomized such that subjects
started with either shoulder internal
or shoulder external rotation exercises.
Within each block, low-,
medium-, and high-intensity rotation
exercises were performed in random
order. After completion of the first
block of exercises, the resting EMG
activity was recorded again for 5 seconds
before the second block of 3
randomly assigned exercises was
begun.
Data Analysis
The data for each subject were analyzed
with MATLAB, version 6. The
raw EMG signals were high-bandpass
filtered (eighth-order zero-lag
Butterworth filter) at 10 Hz to remove
any direct-current offsets and
movement artifacts. The EMG signals
then were rectified and low-bandpass
filtered (eighth-order zero-lag
Butterworth filter) at 2 Hz. The
MVICs and the baseline EMG signals
were processed in the same manner.
The EMG signals were normalized by
first subtracting the baseline EMG
signals from both the EMG signals
and the MVIC activity levels, then
dividing the EMG signals by the
MVIC activity levels, and finally multiplying
by 100%.