Interference-Term Elimination for Wigner Distribution using the Frequency Domain Adaptive Filter.

Abstract

To evaluate an acoustic system by an objective analytical method based on auditory perception, we have focused on the joint time and frequency representation whose processing is assumed to be analogous to human auditory response. For the analysis of a time-varying acoustic signal, the use of a discrete Wigner distribution, WD, has proved to be suitable. However, application of WD is often complicated by the occurrence of interference terms. Thus we proposed a newly revised WD, called RID (reduced interference distribution), and verified that RID remarkably reduced the interference terms of the auto-WD. In this study, we developed another new method which makes use of the block adaptive filter both to suppress the interference terms and to evaluate them quantitatively. In this method, the least mean squares, LMS, algorithm is applied for frequency domain. Simulated signals of a chirp type and measured acoustic signals of a concert hall are analyzed by the methods of spectrograms, WD, RID and the new method. From a comparison among the analyses, it is concluded that the best result can be obtained from the new method developed in this study.