| 摘要 |
An echo canceller for a teleconferencing device is herein disclosed in which a first attenuator limits the dynamic range of a composite speech-echo input signal. A second attenuator, having a gain equal to that of the first attenuator, alters the gain of the emulated echo signal produced by an AFIR filter. The gains of both the first attenuator and the second attenuator are controlled by a common automatic gain control device, which alters the gains as a function of both the dynamic range of the input composite signal and a feedback signal from the output of the first attenuator. This matched attenuator construction allows the AFIR filter to operate within its dynamic range while also limiting the range of the input speech/echo signal. The echo canceller may further comprise a third attenuator which is disposed in the adaptive feedback loop of the AFIR filter. The third attenuator has a gain inversely proportional to the gains of both the first attenuator and the second attenuator. This third attenuator eliminates the gain components from the adaptive error signal. A fourth attenuator may also be provided which limits the received echo source signal, such that the echo canceller is not overloaded.
|
