抄録
A strictly stationary random noise seldom exists in practical random phenomena such as road traffic noise over a long time interval. Furthermore, noise data in a long time interval have recently been automatically measured in the form of quantized digital amplitude level. Accordingly, in order to predict or estimate the statistics of the nonstationary noises, it will be necessary to find an advanced approach to data analysis by use of a digital computer. In this study, with reference to the actual street noise, a new trial to estimate the level distribution of the noise over a long time interval has especially been considered on the basis of the local statistics of the noise fluctuation in a short time interval. It can also be shown that the present estimation procedure is much more suitable to the noise data observed in the form of quantized levels and to any finite number of actual noise data. The validity of the proposed method has experimentally been confirmed by applying it to the street noise data observed in Hiroshima City.
