Application of polynomial regression modeling to automatic measurement of periods of EMG activity

We have developed a new algorithm for automatic detection and measurement of on/off periods of EMG burst and examined validity and reliability of the measuring technique. Mean EMG amplitude (M) during a semi-stationary state of an EMG data array [EMG] is calculated. Because M was determined to be significantly correlated with g(T(on)) (or g(Tend)) which represent amplitude on a polynomial regression curve g(t) which best-fitted to the [EMG], the estimate ]g(T(on) (or ĝ(T(end)) is calculated by substituting M into a regressive equation f(M) which explains the association between the M and g(T(on)) (or g(T(end)). T(on) and T(end) are human-determined on/off burst times for the [EMG]. The on/off periods of the EMG burst are finally computed as roots of the g(t) when ĝ(T(on)) and ĝ(T(end)) are subtracted from the constant of the polynomial. Application of the current method to the human masticatory muscle activity during chewing revealed that the absolute differences between human- and computer-determined measurements were smaller than 10 ms, and these measurements did not differ significantly. We conclude that the proposed algorithm is useful and effective for automatic detection and measurement of on/off periods of EMG burst.