Abstract
Using music as a test signal enables the acoustic parameters of auditoria to be measured, in use, without disturbing users. However, music is not an ideal signal due to its non-stationary characteristics and uneven excitation across the octave band being analysed. In recent years, several methods have been developed for measuring room acoustic parameters using naturally occurring sounds. Currently, the most useful appear to be either based on experiential learning using machine learning techniques or on a maximum likelihood approach. So far, these methods have been restricted to examining the reverberance of the space either through the EDT or RT. This paper will contrast these methods and examine how successful they are at predicting parameters which relate to clarity and strength.