A New Approach for Blind Source Separation of Convolutive Sources
Wavelet Based Separation Using Shrinkage Function
Versandkostenfrei!
Versandfertig in 6-10 Tagen
32,99 €
inkl. MwSt.
PAYBACK Punkte
16 °P sammeln!
ICA and its variations are used extensively in BSS.Most of the algorithms that are used to separatespeech or music signals utilize ICA in the timefrequency domain. Here ICA is applied in the waveletdomain. Separation of signals is achieved byapplying the ICA algorithm and shrinkage functionsto the wavelet coefficients of the originalmixtures. ICA alone can achieve reasonably goodseparation of artificially convolved sources;however, poor separation quality is experienced forreal world convolutive mixtures. This work presentsa novel post processing technique to deal with thecross talk problem. T...
ICA and its variations are used extensively in BSS.Most of the algorithms that are used to separatespeech or music signals utilize ICA in the timefrequency domain. Here ICA is applied in the waveletdomain. Separation of signals is achieved byapplying the ICA algorithm and shrinkage functionsto the wavelet coefficients of the originalmixtures. ICA alone can achieve reasonably goodseparation of artificially convolved sources;however, poor separation quality is experienced forreal world convolutive mixtures. This work presentsa novel post processing technique to deal with thecross talk problem. The post processor is applied tothe signals separated by the ICA network. A superGaussian form of the PDF is assumed for the dominantsource components. Closed form solutions of theparameters of the PDF are obtained by the MOM. ThePDF of the cross talk components is assumed to be ofa GMM, and the EM method is applied to determine theparameters of the Gaussian mixtures. The algorithmis applied to a real world mixture of music andspeech signals. The results show a significantreduction in the cross talk.
Andere Kunden interessierten sich für
