抄録
It occurs very often in an actual observation that the ramdom noise and/or vibration are measured in a digital level from at discrete time intervals, since the various statistical evaluations and the extraction of statistical information become easier by use of digital computer. In this paper, firstly, the unified explicit expression of the multivariate joint probability function of discrete level type for an arbitrary non-stationary random noise and/or vibration fluctuation is found in the general form of statistical orthogonal or non-orthogonal expansion series. The joint probability function which can be arbitrarily chosen in advance for convenience of the statistical data analysis is taken into the first term, and the effects of the fluctuation pattern and its non-stationary properties of random phenomena on the joint probability distribution form is concretely reflected in the second and higher order expansion coefficients in the above expansion expression. Next, a trial of the statistical treatment to predict the noise level distribution over long time interval is considered on the basis of information of the noise statistics within a short time interval. Finaly, the effectiveness of the theoretical results are confirmed experimentally by applying it to the actually measured data of non-stationary road traffic noise.
