A statistical analysis of the spectral moments used in EMG tests of endurance

Abstract

EMG and bandlimited noise (BLN) signals of comparable spectra were processed to compute the centroid and median frequencies and the high/low ratio for different record lengths, signal amplitudes, and noise bandwidths. The spectral moments of EMG and BLN showed similar statistical properties, independent of record length and amplitude. The signal-to-noise ratios of the moments increased with record length. Relatively short records were sufficient to obtain low coefficients of variations especially for the centroid, which was the most stable and sensitive to changes in the underlying spectra. The high/low ratio was the least stable moment. A change of 10 and 3.9% in the centroid over successive records of overall duration of 8 and 10 s, respectively, was just sufficient to detect a spectral shift at the P less than 0.05 level. The equivalent percents for the high/low ratio were 66 and 32.5, respectively. These thresholds of significance were related to time constants of equivalent exponential trends in the spectra used to study fatigue. This relationship makes the results of the present study applicable to a wide variety of EMG tests of endurance.