Data Separation: Independent Component Analysis
Francisco Chinesta, Elías Cueto, Victor Champaney, Chady Ghnatios, Amine Ammar, Nicolas Hascoët, David González, Icíar Alfaro, Daniele Di Lorenzo, Angelo Pasquale, Dominique Baillargeat · Studies in big data · 2025
We consider multi-dimensional data, representing a linear mixture of the hidden sources $$\textbf{s}$$ . For example each component of $$\textbf{s}$$ , $$s_j$$ , could represent a sound source at a certain time (music, voice, noise, ...) and each component of the data $$\textbf{x}$$ , $$x_k$$ represents the sound reception at different spatial location, at the corresponding time. Thus, each receptor mixes the different emission sources. A different vector $$\textbf{x}_i$$ , can be collected at each time instant $$t_i$$ , from the associated sources $$\textbf{s}_i$$ .