engleski

Blind separation of signal sources

Blind source separation (BSS) is a field developed in signal processing and neural network communities over last 15-20 years. It found numerous applications in science and engineering such as acoustics, biomedical signal analysis, communications, image segmentation and deconvolution, spectroscopy, bioinformatics, chemometrics, etc. The basic static linear blind source separation problem is efficiently solved by means of independent component analysis (ICA) under standard assumptions: sources are statistically independent and non-Gaussian, and column-rank of the unknown basis or mixing matrix equals the unknown number of sources. However, in a number of applications statistical independence assumption does not hold completely. Examples include biomedical data sets such as EEG, fMRI, etc. After reviewing basic ICA algorithms for static BSS problem, we shall present algorithms for blind separation of statistically dependent sources. Novel applications of these methods will be demonstrated in single channel blind image and signal deconvolution, blind separation of the images of human faces, and unsupervised decomposition of low-dimensional multispectral images. Concept of sparse component analysis (SCA) will be briefly described and links between SCA and compressed sensing (CS) will be discussed. Novel results relate to CS in learned basis will be demonstrated.

blind source separation; independent component analysis; dependent component analysis; sparse component analysis; nonlinear component analysis

Rad se odnosi na pozvano predavanje na Politechnic of Turin, Laboratory for Engineering of the Neuromuscular Systems, održano 21.05. 2008.