Blind source seperation of mixed visual and audio signals

In this study, separation and deconvolution of blind source mixtures is considered. These mixtures can be lineer or convolutive. Blind Source Seperation algorithms are applied to lineer mixtures of sources but they are unable to separate and deconvolve convolutive mixtures. Delays, reverbation and multi-path effects must be considered in convolutive mixtures. Blind Deconvolution algorithms can be used to deconvolve these mixtures. In this study, to deconvolve and unmix convolutive mixtures, a frequency domain Blind Deconvolution method is presented. The mixtures used in this work, are real-world mixtures recorded in reverberating environment. Blind Source Separation and Blind Deconvolution algorithms are implemented in MATLAB. In this study, it is also presented that convergence rate of frequency domain method can be improved.Keywords: Blind Source Separation, Blind Deconvolution, Frequency Domain Deconvolution, Linear Mixtures, Convolved Mixtures u