Speech enhancement method based on constraint naive generative adversarial network

Abstract

The invention discloses a speech enhancement method based on a constraint naive generative adversarial network. The method comprises the following steps of 1) performing noise data collection and marking; 2) performing voice framing and windowing; 3) performing amplitude compression; 4) inputting the constraint naive generative adversarial network for training; 5) performing amplitude decompression; 6) performing inverse short-time Fourier transform to generate an enhanced speech. The method has the advantages that by adversarial learning between a generative model and a discriminant model inthe generative adversarial network, the sample generation capability of the generative model is continuously enhanced and finally distribution of clean speech samples is obtained; no any assumption exists for statistical distribution of speeches or noises; and a complex number spectrum mapping method is adopted, so that phase information is added in training samples. According to the method, the problem that speech and noise signal distribution is difficult to estimate is ingeniously solved, the speech intelligibility is improved, and phase distortion is avoided.

Description

technical field [0001] The invention relates to the technical field of speech processing, in particular to a speech enhancement method based on a constrained naive generative confrontation network. Background technique [0002] As the main medium of human communication, speech has already played an important role in the fields of mobile communication and multimedia technology. In the context of the ascendant artificial intelligence, the wide application of speech recognition, voiceprint recognition and other technologies has also put forward higher requirements for the quality of speech signals. However, in actual speech acquisition and dialogue communication scenarios, speech signals are often interfered by various noises, mainly including background noise, channel noise and interference noise. Speech enhancement is an effective technique to solve noise pollution. [0003] There are four main traditional speech enhancement methods: (1) spectral subtraction, which uses the...

AI Technical Summary

Problems solved by technology

[0004] The problem to be solved by the present invention is to provide a speech enhancement method based on constrained naive generative confrontation network, which skillfully solves the problem that the distribution of speech and noise signals is difficult to estimate, helps to improve speech intelligibility, and avoids phase distortion

Images