An examination of the Runs Test, Reverse Arrangements Test, and modified Reverse Arrangements Test for assessing surface EMG signal stationarity

The purpose of this study was to examine the accuracy of the Runs Test, Reverse Arrangements Test, and modified Reverse Arrangements Test for assessing stationarity of surface electromyographic (EMG) signals. Five stationary signals were generated by custom programs written with LabVIEW programming software. These signals consisted of sine waves, sums of sine waves, and sums of sine waves and random noise. The sixth signal was a stationary computer generated surface EMG signal downloaded from the surface EMG for the non-invasive assessment of muscles (SENIAM) project database. There were no changes in the amplitude or frequency contents of the stationary signals over time. Several nonstationary signals were also created, including a nonstationary chirp signal generated with LabVIEW programming software, a nonstationary computer generated surface EMG signal downloaded from the SENIAM project database, and a real surface EMG signal recorded from the biceps brachii during a concentric isokinetic muscle action of the forearm flexors at a velocity of 30 degrees s(-1). Both the stationary and nonstationary signals were tested for stationarity using the Runs Test, Reverse Arrangements Test, and modified Reverse Arrangements Test. The results indicated that each of the three stationarity tests demonstrated at least one form of inaccuracy (i.e. false positive and/or false negative results) in examining the stationarity of the test signals. These findings may reflect the fact that these tests were designed to determine whether or not a signal is random, rather than examine signal stationarity exclusively. Thus, the Runs Test, Reverse Arrangements Test, and modified Reverse Arrangements Test may not be appropriate for assessing stationarity in surface EMG signals.