The effect of background muscle activity on computerized detection of sEMG onset and offset

The performance of two computerized algorithms for the detection of muscle onset and offset was compared. The standard deviation (SD) method, a commonly used algorithm, and the approximated generalized likelihood ratio (AGLR) method, a more recently developed algorithm, were evaluated at different levels of background surface EMG (sEMG) activity. For this purpose, the amplitude ratio between the period of muscle inactivity and activity was varied from 0.125 to I in artificially assembled sEMG traces. In addition, 1230 real sEMG signals, obtained from various trunk muscles during quick release, were raised to a power of 3 to change their relative amplitude ratio. As the relative level of background activity increased, both the SD and AGLR methods produced longer latencies and detected fewer muscle responses, suggesting that a detection artifact can be introduced if the subject populations being compared have different levels of background muscle activity. Of the two methods, AGLR appears to be the least affected by background activity. However, above the ratio 0.8, results from AGLR are also unreliable, particularly in detecting offsets. Average latency artifacts near this ratio were 8 ms for AGLR and 46 ms for SD. (c) 2007 Elsevier Ltd. All rights reserved.