Linear geometric ICA: fundamentals and algorithms
Neural Computation
A unifying model for blind separation of independent sources
Signal Processing
IEEE Transactions on Signal Processing
Fast and robust fixed-point algorithms for independent component analysis
IEEE Transactions on Neural Networks
Sparse component analysis and blind source separation of underdetermined mixtures
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
This paper addresses the blind source separation (BSS) problem where the input data are mixtures of sources that are ''sparse''-that is each source has zero amplitude for some of the time. It is shown that, under certain conditions, it is possible to separate these sources by the analysis of localised segments of the phase space trajectory where one source dominates rather than applying statistical methods to the whole phase space plot. Results are presented for both simulated data, and thoracic and abdominal ECG data taken from an expectant mother. It is shown that when applied to the ECG data, the proposed technique has a comparable performance to the standard Fast ICA method. For the particular simulated data set, the proposed method gives better results for relatively low noise levels but is less robust than the FastICA for higher noise levels. Subsequently, the potential advantage of using the proposed technique for mixtures of correlated sparse sources is demonstrated. However, for more general sources that are not sparse, the proposed technique in its present form has, as expected, an inferior performance compared to the FastICA technique.