Evaluation of Cross-talk in Electromyographic Signals

Abstract

Activity of skeletal muscles produce electrical signals that can be measured using electrodes placed on the skin surface over a target muscle or with electrodes inserted into the muscle. Such electromyographic (EMG) signals provide fundamental information about the intensity of the neural drive acting upon muscle. In addition, EMG signals are widely deployed as control sources for powered prosthetic limbs. One limitation related to recording EMG signals, however, is that signals arising from neighboring muscles may contribute significantly to the activity detected with electrodes placed over or within a given target muscle. Such unwanted contribution of signal from muscles other than the targeted muscle is referred to as cross-talk. Cross-talk was investigated in four neighboring muscles in the forearm with different electrode types and configurations: bipolar intramuscular, monopolar intramuscular, and bipolar surface EMG. Cross-correlation analysis was performed for every pairwise combination of EMG signal recorded. The peak correlation coefficient at near-zero time delay provided an index of the degree of cross-talk. Correlation coefficients dropped off exponentially with distance between recording electrodes. Bipolar intramuscular EMG had the narrowest pick-up range, with a length constant of 14.5 mm. Bipolar surface EMG had a longer length constant of 37.0 mm, whereas monopolar intramuscular EMG had the longest length constant of 64.5 mm. A second set of experiments indicated that correlation in EMG signals detected in different muscles was unlikely to have a neural basis. Therefore, because of their wide detection range, monopolar configurations including those involving intramuscular electrodes, should be avoided.