A DATA DEPENDENT COMPUTER ALGORITHM FOR THE DETECTION OF MUSCLE-ACTIVITY ONSET AND OFFSET FROM EMG RECORDINGS

This paper describes modifications to an algorithm presented by Marple -Horvat and Gilbey (1992) for identifying bursts of muscle activity in electromyographical (EMG) recordings. Our efforts to apply their algo rithm to spontaneously moving infants and toddlers resulted in limited success. The modified algorithm makes several parameters dependent on the data being analyzed; these changes enabled it to analyze a variet y of EMG recordings more effectively. The original algorithm had a suc cess rate (correctly identified bursts) of 62.9% and combined error ra te (number of insertions and deletions) of 73.0% when applied to an in dependent test data set. The modified algorithm displayed a success ra te of 85.4% and combined error rate of 23.6%. (C) 1998 Elsevier Scienc e Ireland Ltd.