Distortion compensation in an acoustic echo canceler
View Patent ↗An audio communications system has an acoustic echo cancellation (AEC) module. The AEC module receives a digital signal sent to a loudspeaker and a digital signal received from a microphone. The signal received from the microphone contains an echo of the signal played through the loudspeaker. The loudspeaker signal is processed by an audio generation module (AGM) that models substantially nonlinear distortions that can occur while producing the signal played through the loudspeaker. The AGM includes a modeling path comprised of one or more distortion modules. Each distortion module receives digital samples as input, modifies the samples to model a form of distortion, and outputs the modified samples. The output of the AGM is provided to an acoustic echo estimation (AEE) module that uses adaptive algorithms to compensate for substantially linear changes in the echo characteristics of the environment in which the loudspeaker and microphone are located.
1. An audio communication device, comprising:
a first audio signal input for receiving a first audio signal;
a second audio signal input for receiving a second audio signal, wherein at least a portion of the second audio signal is an echo of the first audio signal;
an echo cancellation device, including
a subtractor having two inputs and one output, and configured to subtract signals presented on the two inputs to remove the echo and to output a subtracted signal on the subtractor output, the first subtractor input being connected to the first audio signal output,
a noise reducer having an input connected to the subtractor output, and having an output,
an auto gain control having an input connected to the noise reducer output,
an audio sensing filter having an output connected to the subtractor second input, and having an input, the audio sensing module configured to compensate for distortions caused by sensing the first audio signal,
an acoustic echo estimation filter having an output connected to the audio sensing module input, and having an output, said acoustic echo estimator configured to compensate for substantially linear changes in echo characteristics, and
an audio generation filter having an output connected to the acoustic echo estimator input, and having an input connected to the second audio signal input, the audio generator configured to remove substantially nonlinear distortions associated with the formation of the first audio signal.
2. The device of claim 1 , wherein the audio generation filter comprises:
an amplifier clipper;
a voice coil displacement filter; and
a harmonic distortion filter.
3. A method of canceling an echo in an audio signal, comprising the steps of
receiving a first audio signal;
receiving a second audio signal, wherein at least a portion of the second audio signal is an echo of the first audio signal; and
cancelling said echo, including
subtracting signals presented on two inputs to remove the echo and to output a subtracted signal, the signal on one of said two inputs being related to the first audio signal, and the signal on the other of said two inputs being related to said second audio signal,
reducing noise in said subtracted signal,
performing auto gain control on the noise reduced subtracted signal, and outputting the gain controlled, noise reduced subtracted signal,
modifying digital samples corresponding to said second audio signal to remove substantially nonlinear distortions associated with the formation of the first audio signal, and outputting a first modified signal,
modifying said first modified signal to compensate for substantially linear changes in echo characteristics to produce a second modified modeled signal,
modifying the second modified modeled signal to compensate for distortions caused by sensing the first audio signal, and outputting a third modified signal as said signal on the other of said two inputs being related to said second audio signal.
4. The method of claim 3 , wherein the step of modifying digital samples corresponding to said second audio signal to remove substantially nonlinear distortions comprises the step of:
filtering with an amplifier clipper;
filtering with a voice coil displacement filter; and
filtering with a harmonic distortion filter.