Classification of room impulse responses with self-organizing maps

Abstract

In this paper, a method for classifying room impulse responses using multifractals and Kohonen's neural networks is investigated. Impulse response is basic source of information in room acoustics; therefore its analysis is the most important issue regarding sound impression in the room. The method proposed in this paper consists of three steps. The first stage of signal classification process is computation multifractal spectrum of the signal. Main features of multifractal spectrum are extracted in the second step. Grouping of similar signals based on extracted features is done in the third step. For every group of signals formed in previous step, model of desirable multifractal spectrum is determined. The experimental results verify the usability of described algorithm.