Input signal-based frequency domain adaptive filter stability control

摘要

An adaptive filter converts time domain samples of an input signal into frequency domain signals, dynamically adjusts a stability conditional number based on the frequency domain signals, and uses the dynamically adjusted stability conditional number to control step size normalization during adaptation of frequency domain coefficients of the adaptive filter. The stability control number may be global to a range of frequency bins based on a peak magnitude of the input signal and/or may be frequency bin-specific stability control numbers based on corresponding frequency bin-specific error signal magnitudes. The adaptive filter also dynamically adjusts a noise floor based on the frequency domain input signals and refrains from updating frequency domain coefficients when a magnitude of the frequency domain input signal associated with a frequency bin is greater than the dynamically adjusted noise floor.