1985 年 30 巻 1 号 p. 55-63
Since EMG waveform detected by surface electrodes is a complex spatio-temporal summation of many muscle fiber potentials, it is considered as a random signal or a stochastic process. Statistical methods can play a useful role in estimating the physiological informations from such signals. The present study placed focus on a myoelectric activity during a sustained contraction which was one of the fundamental movements in man. In order to establish the basis of statistical analysis of the EMG, the normality and stationarity which were the most basic statistical properties of the stochastic process were strictly examined. The experimental and processed results showed that the probability density function of the EMG amplitude was the first order normal distribution. The mean was not affected by a shift of time origin and remained in a constant level, i.e., zero. The autocorrelation function depended only on the time lag (τ) and had a positive finite value only for τ=0. From these features, it was verified that the EMG was stationary in wide sense; Furthermore, its important properties of uncorrelated and orthogonal nature were revealed.