I am currently analyzing some EMG data from a study looking at
changes in muscle activity over learning
time in unconstrained, multi-joint movements. In an effort to
quantify timing changes in muscle activity patterns, I am using
threshold and duration criteria to turn the EMG signals into "on/off"
signals for each muscle. Thresholds are computed for each muscle

In order to determine the threshold I am currently using, I collected
2 "baseline" trials before and after my learning session for each
subject. During the baseline trials, the subjects were in a resting
position. Thus, the EMG shows no activity. I'm extracting an average
max voltage across 1 second bins for these baseline trials and using
this value as my threshold criterion.

The strategy above works well when analyzing both tonic and phasic
EMG activity. Nevertheless, I'm also interested in trying to look at
more phasic activity to the exclusion of tonic activity. The only
folks I know of that do this type of thing are Martha Flanders and
colleagues up at Minnesota. However, the technique they use to
extract phasic activity doesn't seem to be very applicable to a
learning study.

One way of looking at the phasic activity might be to use multiple
thresholds. I have thought about taking a proportion of the max
voltage for each muscle and using that as my threshold. However, I'm
not sure if this technique is commonly used in other EMG research.

Thus, my questions are as follows:
1) Is it common for folks to use a proportion of max EMG voltage as a
threshold criterion and, if so,
2) What proportion values are commonly used (I was planning to use
10 and .20...)?

Thanks in advance,

John Spencer
Dept. of Psychology
Indiana University