Noisy speech signal processing method, device and server

Abstract

The invention discloses a noisy speech signal processing method, device and server, belonging to the technical field of communication. The method includes: according to the silent section of the noisy speech signal, obtaining a noise signal in the noisy speech signal; for each frame in the speech signal, obtaining each frame of the speech signal according to the noise signal and the noisy speech signal The power spectrum iteration factor of a frame; according to the power spectrum iteration factor of the noisy speech signal, each frame of the noise signal and the previous frame, calculate the intermediate power spectrum of each frame of the speech signal; according to each frame of the speech signal The intermediate power spectrum and the noise signal, calculate the signal-to-noise ratio of each frame in the noisy speech signal; according to the signal-to-noise ratio of each frame in the noisy speech signal, each of the noisy speech signal and the noise signal frame to obtain the processed noisy speech signal in the time domain. The invention processes the noisy speech signal through the power spectrum iteration factor, thereby improving the auditory quality of the user.

Description

technical field [0001] The present invention relates to the field of communication technologies, in particular to a method, device and server for processing noisy speech signals. Background technique [0002] Speech in real life is inevitably affected by ambient noise. In order to improve the auditory quality, it is necessary to denoise the speech signal. [0003] In the denoising process, an algorithm based on short-term amplitude spectrum estimation is usually used, that is, in the frequency domain, the power spectrum of the speech signal is obtained by using the power spectrum of the original speech signal and the power spectrum of the noise signal, and according to the power spectrum of the speech signal The amplitude spectrum of the speech signal is calculated, and the speech signal in the time domain is obtained through inverse Fourier transform. [0004] In the process of realizing the present invention, the inventor finds that there are at least the following proble...

AI Technical Summary

Problems solved by technology

[0005] For signal power spectrum estimation, the usual method is to use an iterative algorithm with a fixed iteration factor. This algorithm is often effective for white noise and cannot track changes in speech or noise in time, so the performance drops sharply when encountering colored noise.

Images