ICA Based Feature Extraction from One-Dimensional Signal for Machine Condition Monitoring
Qingbo He, Ruxu Du, Fanrang Kong · 2008
This paper proposed a new feature extraction method based on independent component analysis (ICA) from one- dimensional signal. The ICA based feature corresponds to the higher-order statistics. It contains plentiful phase information and thus, has the merit in some applications. The new feature extraction is done in three steps: first, the ICA basis filters of one class signal are trained by a number of short segments of the signal; the measured signals are then sent to the ICA basis filters to get the transformed coefficients; finally, a new feature called ICA filtered correlation feature is quantitatively calculated by the transformed coefficients. The new feature has the clear class property and can be applied for signal classification. The experimental verification shows the effectiveness of the new feature and the value for machine condition monitoring.