Automatic equalization using adaptive frequency-domain filtering and dynamic fast convolution

摘要

Frequency-domain techniques are used for adaptive equalization that is responsive to spectral magnitude characteristics but not sensitive to phase characteristics of system response. Signal correlation may be used to improve adaptation accuracy when significant levels of ambient sounds are present. A preferred filter implementation uses convolution-based block transforms and cross-fade windows.

主权项

1. A method for equalizing an audio signal, wherein the method comprises:
receiving the audio signal; applying an adaptive filter to the audio signal to generate a filtered audio signal; driving an acoustic output transducer with the filtered audio signal to generate a sound field; generating a detection audio signal by detecting the sound field after propagating through an acoustic channel with an acoustic input transducer; obtaining a first frequency-domain representation from the detection audio signal that is a composite spectral magnitude response of the acoustic output transducer, the acoustic channel and the acoustic input transducer; applying a delay to the filtered audio signal to obtain a delayed filtered audio signal, wherein the delay includes an approximation of propagation delays of the sound field through the acoustic channel; obtaining a second frequency-domain representation from the delayed filtered audio signal; processing the first frequency-domain representation, the second frequency-domain representation, and a third frequency-domain representation from spectral components of a target spectral magnitude response to generate an estimated spectral magnitude response correction signal; obtaining an equalization-filter control signal from the estimated spectral magnitude response correction signal; and configuring adapting filter characteristics of the adaptive filter in response to the equalization-filter control signal.