| 摘要 |
A system for estimating delay between a far-end signal and an echo of the far-end signal in a microphone signal, the system comprising: a buffer configured to store a time-series of far-end samples representing the far-end signal; a first delay estimator configured to form a first estimate of the delay in respect of a speech frame representing speech in the microphone signal; a second delay estimator configured to form a second estimate of the delay for the speech frame by operating a first set of one or more filters on far-end samples selected from the buffer in dependence on an operating delay; a second set of one or more filters for operation on far-end samples; and a controller configured to, in response to a determination that the first estimate of the delay for the speech frame differs from the operating delay for a previous frame by at least a predefined length of time, cause the one or more filters of the second set to operate on far-end samples selected from the buffer according to the first estimate of the delay and, if in respect of one or more speech frames a measure of convergence of the second set of filters exceeds a measure of convergence of the first set of filters by at least a first predefined threshold, update the operating delay using the first estimate of the delay. |
