Method and device for de-reverberation of audio signal

Abstract

Examples of the present invention disclose a method and device of an audio signal.Among them, get the first audio reverberation signal collected from the audio channel and the second audio reverberation signal. The first audio reverberation signal is obtained at the current frame to M frequency point.The frame -to -M frequency point is collected. According to the n -frequency point in the M frequency point according to the first audio reverberation signal and the second audio reverberation signal, the return coefficient of the room corresponding to the historic frame is updatedThe return coefficient of the room corresponding to the current frame is configured to configure the known value to the (m‑n) frequency point in the current frame in the room corresponding to the room of the current frame.According to the room regression coefficient corresponding to the current frame, the pure voice signal corresponding to the current frame is obtained.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and device for de-reverberating audio signals. Background technique [0002] In far-field voice interaction, the microphone is usually far away from the sound source, and the received audio signal will be affected by reverberation. Reverberation greatly reduces the recognition accuracy of audio signals, leading to a sharp decline in the accuracy of existing speech recognition. [0003] In order to reduce the influence of reverberation on the audio signal, the received audio signal is usually de-reverberated. De-reverberation can be improved to a certain extent to enhance the experience of voice interaction. In the prior art, there are some dereverberation algorithms for filtering out reverberation in audio signals. However, these de-reverberation algorithms usually have the following problem: when solving the filter coefficients for filtering reverberation, a large ...

AI Technical Summary

Problems solved by technology

However, these reverberation algorithms usually have the following problems: when solving the filter coefficients for filtering reverberation, a large number of calculations of matrix inversion and matrix multiplication are often involved

The amount of calculation between the matrices is large and the algorithm complexity is high, so the real-time performance of the reverberation algorithm will be poor, which will affect the performance of understanding reverberation

Images