Improving the Robustness of the Correlation Approach for Solving the Permutation Problem in the Convolutive Blind Source Separation

n this paper, we propose a modification to thecorrelation approach in convolutive blind source separationto achieve an improved robustness. An often used approachfor separation of convolutive mixtures is the transformationto the time-frequency domain. This allows for the use of aninstantaneous ICA algorithm independently in each frequencybin, which greatly reduces complexity. The drawback of thisapproach are the so-called permutation and scaling problems.Here, we modify the well known correlation approach for makingit more robust. We propose to incorporate a confidence functionbased on estimated SIR which allows for detection of frequencybins with high probability of wrong permutations. The results ofthe new algorithm will be shown on an real-world example.