Effects of weight load on physiological tremor: the AR representation.

The propriety of the autoregressive (AR) method as a means of processing the signals of physiological tremor of human finger (finger tremor) was investigated. Application of the Akaike's criterion demonstrated that the 15-th order AR model was required to describe the recordings of finger tremor. According to Burg's algorithm, both AR spectrum and AR parameters were estimated to study the effects of various weight loads on finger tremor. It was found that, (1) the amplitude of AR spectrum was apparently enhanced by adding the load; (2) the first prediction coefficient (a1) and the first reflection coefficient (rho 1) significantly declined by increasing the weight loads. The results were compared with the calculations from FFT (Fast Fourier Transform) and autocorrelation function. Simple physical interpretation of the AR parameters (i.e. a1 and rho 1) was discussed in relation with system's resonant modes.