Application of Independent Component Analysis to Decomposition of Engine Acoustic Signals
Liu Yue-hui · Journal of Tianjin University Science and Technology · 2006
Independent component analysis (ICA) was carried out to decompose engine noise signals. Based on FastICA algorithm, the conventional ICA model for engine noise was set up, and further, the sequential independent component analysis model was applied. Take a 4-cylinder diesel engine for example, the recorded acoustic signals were computed for kurtosis, and found to be non-Gaussian. So the acoustic signals were then decomposed into several independent components by sequential independent component analysis. And the wavelet transform was applied to these components to get their time-frequency information. It was shown that these independent components were the approximation of different engine noise source signals.