It looks like there's a
difference in emphasis between people who do feature extraction where a feature signal is presumed to
exist and suppression of noise is good, and those who do hypothesis testing, where you are asking whether a feature is provably present, and you need to compare the variance due to signal with that due to noise.