Continuous time-varying autoregressive spectrum for assessment of fatigue induced changes in SEMG signals

The change of frequency spectrum has been used as an indication of a muscle fatigue process for more than two decades. Mainly FFT and linear modeling in combination with the sliding window approach have been used for estimation of the time-varying spectrum. Our proposal in this paper is that the technique in which the time-varying coefficients of the linear model are assumed to be a linear combination of the basis functions is advantageous in some applications, because it produces a smooth, continuous spectrum providing a better view to the slowly changing frequency content of a wide-band non-stationary signal.