Asynchronous clock frequency domain acoustic echo canceller

摘要

An echo cancellation system that detects and compensations for differences in sample rates between the echo cancellation system and a set of wireless speakers based on a frequency-domain analysis of estimated impulse response coefficients. The system tracks the real and imaginary number components of the coefficients, and determines a “rotation” of the coefficients over time caused by a frequency offset between the audio sent to the speakers and the audio received from a microphone. Based on the rotation, samples of the audio are added or dropped when echo cancellation is performed, compensating for the frequency offset.

主权项

1. A method, comprising:
transmitting a constant sinusoidal tone to a first wireless speaker as a first reference signal; receiving a first signal from a first microphone, the first signal including audible sound output by the first wireless speaker; applying a Fast Fourier Transform (FFT) to the first signal to determine a first frequency domain signal; applying the FFT to the first reference signal to determine a first frequency domain reference signal; filtering the first frequency domain reference signal using an adaptive filter; subtracting an output of the adaptive filter from the first frequency domain signal to determine a first frequency domain output signal; iteratively calculating a first frequency domain estimated impulse response coefficient of the adaptive filter based on the first frequency domain output signal; determining a first angle, the first angle being that of a first vector of a first iteration of the first frequency domain estimated impulse response coefficient relative to a real number axis and an imaginary number axis, the first vector corresponding to a first real number component and a first imaginary number component of the first iteration; determining a second angle, the second angle being that of a second vector of a second iteration of the first frequency domain estimated impulse response coefficient relative to the real number axis and the imaginary number axis, the second vector corresponding to a second real number component and a second imaginary number component of the second iteration; performing a first linear regression to determine a first linear fit based on the first angle and the second angle; determining a first frequency offset between the first reference signal and the first signal based on the first linear fit, wherein the first frequency offset is a difference between a first sampling rate of the first reference signal and a second sampling rate of the first signal; determining that the first frequency offset is negative; and skipping at least one sample of the first reference signal prior to applying the FFT to the first reference signal to eliminate the first frequency offset.