Presentation 2015-03-02
Estimating correlation coefficients of two super-Gaussian complex signals without phase observation
Shigeki MIYABE, Nobutaka ONO, Shoji MAKINO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we describe estimation of a correlation coefficient between two complex signal sequences under the condition where the observation misses the phase. In our previous work, we formulated a probabilistic model which assumes that the complex amplitude sequences follow a bivariate complex normal distribution, and proposed a maximum likelihood estimation of the correlation coefficient by an EM algorithm which treats the phase difference as a hidden variable. However, complex signals are often super Gaussian and cause model mismatch of the Gaussian assumption, and the estimation accuracy depends on signals. In this paper, we examine an estimation robust against the model mismatch by formulating a maximum likelihood estimation adaptive to the signal shapes by assuming that the two complex amplitude sequences follow a multivariate t distribution. Experimental results reveals that the complex t distribution model is not always better than the complex normal distribution model depending on the signals, but by selecting the appropriate model, the maximum likelihood estimation can obtain the better result than a straightforward amplitude correlation estimator.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) correlation coefficient / maximum likelihood estimation / EM algorithm / super Gaussianity
Paper # EA2014-75,SIP2014-116,SP2014-138
Date of Issue

Conference Information
Committee EA
Conference Date 2015/2/23(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Engineering Acoustics (EA)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Estimating correlation coefficients of two super-Gaussian complex signals without phase observation
Sub Title (in English)
Keyword(1) correlation coefficient
Keyword(2) maximum likelihood estimation
Keyword(3) EM algorithm
Keyword(4) super Gaussianity
1st Author's Name Shigeki MIYABE
1st Author's Affiliation Life Science Center of Tsukuba Advanced Research Alliance, University of Tsukuba()
2nd Author's Name Nobutaka ONO
2nd Author's Affiliation Principles of Informatics Research Division, National Institute of Informatics:School of Multidisciplinary Sciences, The Graduate University for Advanced Studies (SOKENDAI)
3rd Author's Name Shoji MAKINO
3rd Author's Affiliation Life Science Center of Tsukuba Advanced Research Alliance, University of Tsukuba
Date 2015-03-02
Paper # EA2014-75,SIP2014-116,SP2014-138
Volume (vol) vol.114
Number (no) 473
Page pp.pp.-
#Pages 6
Date of Issue