Konferenční příspěvek

Koldovský Z., Málek J., Tichavský P., Deville Y., Hosseini S

ICASSP 2008: IEEE International Conference on Acoustics, Speech, and Signal Processing

Conference Management Services

Bryan

2008

1-4244-1484-9

Ing. Petr Tichavský, CSc. (ÚTIA - Oddělení stochastické informatiky)

independent component analysis, piecewise stationary signals

We propose an extension of algorithm EFICA for piecewise stationary and non-Gaussian signals. Cramér-Rao bound for this model is derived and the extended algorithm is shown to be asymptotically efficient if score functions of the signals are known in each block. In contrast to classical ICA algorithms, the proposed method is able to profit from varying distribution of the original signals as well as from their non-stationarity, which is demonstrated on simulations with real-world signals.