Blending Independent Components and Principal Components Analysis

2.1 What does ICA aim to achieve?

[this page | pdf | references | back links]

Return to Abstract and Contents

Next page


2.1          What does ICA aim to achieve?


Independent component analysis (ICA) is a tool for extracting useful information from a large amount of data. It seeks to identify the driving forces that underlie a set of observed phenomena. The phenomena to which ICA could be applied are very wide ranging, including mobile phone signals, stock price returns, brain imaging or voice recognition. It belongs to a class of blind source separation (BSS) methods for separating data into underlying informational elements, where ‘blind’ here means that such methods endeavour to separate data into source signals even if very little is known about the nature of the source signals.


Generally speaking, ICA is applied in a situation where there are several distinct ‘output’ signals able to be measured, and it is reasonable to postulate that these outputs depend on combinations of distinct underlying ‘input’ signals or factors. The aim is, as far as possible, to estimate or characterise these underlying signals. For example, with voice recognition, the aim might be to differentiate between different foreground contributors to the overall sound pattern and, additionally, to filter out, as far as possible, anything that appears to be background noise.


Contents | Prev | Next

Desktop view | Switch to Mobile