ABSTRACT
Independent component analysis (ICA) achieves decomposition of the observed data matrix based on the assumption of statistical independence among the components. The state-of-the-art performance measures to measure the quality of ICA algorithms are based on the source time sequence matching criteria. The article proposes three new classes of performance measures. They are based on statistical matching of the estimated and actual sources, the value of the independence measures achieved and the separation quality achieved in comparison with the best, where the best can be obtained using the performance measures themselves used as independence measures.
