Decorrelating MVDR Filterbanks Using the Non-Uniform Discrete Fourier Transform | IEEE Journals & Magazine | IEEE Xplore

Decorrelating MVDR Filterbanks Using the Non-Uniform Discrete Fourier Transform


Abstract:

Minimum variance distortionless response (MVDR) is a classic design criteria in signal adaptive spectral analysis as well as filter design. We extend this approach to fil...Show More

Abstract:

Minimum variance distortionless response (MVDR) is a classic design criteria in signal adaptive spectral analysis as well as filter design. We extend this approach to filterbanks with the constraint that transform domain signal components must be uncorrelated. Our analysis shows that filterbanks based on Vandermonde decomposition of the autocorrelation matrix correspond to the non-uniform discrete Fourier transform and satisfies the MVDR criteria. Namely, the columns of the inverse Vandermonde matrix corresponds to filters with unit response at the pass-band while leakage is minimized with the constraint that components remain uncorrelated. In the special case that the autocorrelation matrix is rank deficient, the proposed filterbank coincides with Pisarenko's harmonic decomposition, which thus also satisfies the MVDR criteria.
Published in: IEEE Signal Processing Letters ( Volume: 22, Issue: 4, April 2015)
Page(s): 479 - 483
Date of Publication: 16 October 2014

ISSN Information:

International Audio Laboratories Erlangen, Fraunhofer Institute of Integrated Circuits, Erlangen, Germany

I. Introduction

Signal processing algorithms, especially within speech and audio, rely frequently on filterbanks which provide a time-frequency representation of the signal [1]. Such representations are attractive for three main reasons. Firstly, they give access to components of the signal which carry a physically interpretation, enabling easy application of physically motivated methods. Secondly, since many methods, such as the short-time Fourier transform, give frequency components which are approximately uncorrelated, we can process frequency components independently from each other. This allows design of efficient processing algorithms since we do not have to take cross-correlations into account. Thirdly, to compute time-frequency transforms, we can use superfast algorithms such as the fast Fourier transform (FFT).

International Audio Laboratories Erlangen, Fraunhofer Institute of Integrated Circuits, Erlangen, Germany

References

References is not available for this document.