Method and device for noise enhancement post-processing in speech decoding

Abstract

The embodiment of the invention discloses a method for noise enhancement post-processing in speech decoding. The method comprises the following steps: the noise factor is obtained according to the spectral envelope and the long-time mean spectral envelope of a speech frame; whether the noise factor is higher than the preset threshold value or not is judged; the energy of the fixed code vector of the speech frame is smoothed according to the noise factor and the adaptive codebook gain of the speech frame; and the smoothed fixed code vector is obtained according to the energy of the smoothed fixed code vector. The embodiment of the invention also discloses a noise enhancement post-processing device. By adopting the method and the device, the complexity of calculation of the noise enhancement post-processing in the speech decoding can be lowered.

Description

technical field [0001] The invention relates to the communication field, in particular to a noise enhancement post-processing method and device in speech decoding. Background technique [0002] When a communication system transmits a voice signal, in order to effectively compress the voice signal to make it suitable for transmission and storage, the voice signal needs to be encoded and decoded. For example, in a digital system, the speech signal is first converted into a digital speech signal, and then the speech encoder encodes the digital speech signal to obtain a compressed bit stream; the compressed bit stream is transmitted through the channel, and at the receiving end, the speech decoder compresses the bit stream. The bit stream is reconstructed into a speech signal. [0003] In code-excited linear prediction (Code Excited Linear Prediction, CELP) coding, digital speech signals are processed in units of frames. For each frame of digital audio signal, the encoder extr...

AI Technical Summary

Problems solved by technology

[0011] In the process of realizing the present invention, the inventor found that the judgment of noise in the above-mentioned technology is more complicated (first judge voice / non-speech, and then judge stable noise / non-stable noise, this judgment process is more cumbersome, and the calculation complexity is higher), At the same time, in the process of energy smoothing and ISP coefficient smoothing of noise, the smoothing complexity is high, which reduces the efficiency of post-processing of noise enhancement.

Images