Described is an echo cancellation method including buffering the at least one frame of a first voice signal, the at least one frame of the first voice signal including an echo signal, when the at least one frame of signal is preprocessed, separately obtaining, in a first timing period, the at least one frame of the first voice signal from a buffer and at least one frame of a reference signal matching the at least one frame of first voice signal from a reference signal queue, so that an estimated latency value, between each frame of first voice signal and a corresponding reference signal, remains within a preset range, performing echo cancellation processing on the echo signal in the at least one frame of first voice signal using the at least one frame of reference signal, to obtain a second voice signal, and outputting the second voice signal.