Choosing a non-Rayleigh reverberation model
Douglas A. Abraham · 1999
Two methods for choosing a non-Rayleigh reverberation model are presented, a graphical method based on higher order moments and a more quantitative method based on multiple hypothesis testing. The latter method first considers two non-parametric tests which are subsequently seen to be inferior to a generalized likelihood ratio test (GLRT) in terms of minimizing an average log/sub 10/ P/sub fa/ error measure. However, once the GLRT has designated a model, the non-parametric methods may be used to obtain a measure of the confidence of the decision through evaluation of their p-values. It is also observed that the Rayleigh mixture model is used most frequently (of the models considered); that is, the Rayleigh mixture model was most robust in representing reverberation from a variety of source models (excepting log-normal).