Echo cancellation model training method and device and echo cancellation method and device

Abstract

The invention relates to an echo cancellation model training method and device and an echo cancellation method and device, and the training method comprises the steps: obtaining a first signal according to a near-end signal; inputting the first signal into an echo cancellation model, and obtaining a second signal based on the output of the echo cancellation model, the second signal being a prediction signal of the first signal after echo cancellation; according to the first signal and the second signal, prediction loss of an echo cancellation model is determined, the prediction loss comprises first prediction loss, and the first prediction loss is calculated based on a residual echo component and a voice signal component separated from the second signal; and adjusting parameters of the echo cancellation model according to the prediction loss. According to the training method and device of the echo cancellation model and the echo cancellation method and device disclosed by the invention, the residual echo can be additionally suppressed, and the echo cancellation amount is improved.

Description

technical field [0001] The present disclosure relates to the field of signal processing, and in particular, to a method and device for training an echo cancellation model and a method and device for echo cancellation. Background technique [0002] When conducting real-time communication with far-end users indoors, the sound from the far-end can be played on the near-end speaker, reflected and propagated in the indoor space, and then re-collected by the near-end microphone to form an acoustic echo. Acoustic Echo Cancellation (AEC) can identify and suppress echo signals, so that remote users cannot hear their own echoes, which greatly improves the audio experience. In the related art, echo cancellation can be performed by AEC incorporating a deep learning method, but the amount of echo cancellation is limited, and the phenomenon of echo leakage is easy to occur. SUMMARY OF THE INVENTION [0003] The present disclosure provides a method and device for training an echo cancel...

AI Technical Summary

Problems solved by technology

In related technologies, echo cancellation can be performed by AEC integrated with deep learning methods, but the amount of echo cancellation is limited, and the phenomenon of echo leakage is prone to occur.

Images