Table 7.

Comparison of feature extraction methods.

Method	Advantages	Disadvantages
AR	- Good frequency resolution	- Low performance when applied to non-stationary signals
	- Limited spectral loss	- The order of the model is difficult to select
FFT	- Good performance when applied to stationary signals	- High noise sensitivity
	- Appropriate for narrowband signals	- Poor performance on non-stationary signals
	- Good speed for real-time applications	- Weak spectral estimation
WT	- Varying window size to analyze several frequencies	- Proper mother wavelet selection is not trivial
	- Good to analyze transient signal changes