抄録
As one of evaluation methods for the room sound fluctuation, a new trial to predict the probability distribution of the response sound fluctuation in the room excited by arbitrary type environmental noise input is proposed by emphasizing its methodological viewpoint. The response sound intensity of the room can be naturally expressed as a total sum on intensity scale of each frequency response component and its modelling error caused by the quantized error and/or the observation noise. Owing to the non-Gaussian property of intensity variable and modelling error, the well-known system identification method is too difficult to be directly applied to the present case. So, in this paper, first by newly introducing some kind of information measure in a form of mutual entropy quantity between input and output and maximizing it, a new estimation algorithm of acoustic characteristic parameters of the room is derived by applying a stochastic approximation method to the above extremal condition with using a general expression form of the Gram-Charier series expansion type non-Gaussian probability distribution. In the room excited by an arbitrary environmental noise input fluctuating in non-Gaussian type probability distribution form, the probability distribution of response sound fluctuation for an overall frequency range and each frequency band component can be predicted by using the above identified system model. The proposed method is applied to the room simulating the actual living acoustic situation and its justification and usefulness is experimentally confirmed, too.
