Voice enhancement method under low signal-to-noise ratio condition based on voiceprint embedding

Abstract

The invention provides a voice enhancement method under a low signal-to-noise ratio conditionbased on voiceprint embedding. The method comprises the steps: making a clean Chinese speech data set and a Chinese speech noise data set mixed with clean speech and random noise through sox to obtain a noisy speech data set; extracting a Mel-frequency cepstral coefficient of the Chinese speech data set; constructing and optimizing a universal Gaussian mixture background model; according to the Mel-frequency cepstral coefficient of the clean Chinese speech, optimizing a Gaussian mixture background probability density model, and extracting voiceprint features of the clean Chinese speech data set; extracting an amplitude spectrum and a phase spectrum of the noisy voice data set; generating related spectrum features of a speaking user in the noisy voice; constructing an enhanced neural network; and performing waveform reconstruction to obtain enhanced noisy voice. According to the invention, the voiceprint features of the user voice are embedded, the performance of the voice enhancement system under the condition of low signal-to-noise ratio is improved, and the performance of the intelligent voice equipment in a noisy environment is improved.

Description

technical field [0001] The invention relates to the field of speech enhancement, in particular to a speech enhancement method under the condition of low signal-to-noise ratio based on voiceprint embedding. Background technique [0002] In recent years, the popularity of artificial intelligence technology has remained high, and speech enhancement technology has also developed rapidly, and various speech enhancement technologies emerge in endlessly. These speech enhancement schemes are mainly divided into: traditional speech enhancement schemes and deep learning-based speech enhancement schemes. [0003] Traditional speech enhancement schemes mainly include: spectral subtraction, statistical model-based enhancement algorithms and subspace enhancement algorithms. Spectral subtraction assumes that the noise is additive and then subtracts an estimate of the noise spectrum from the speech spectrum of the noisy speech, resulting in clean speech. The Wiener filter algorithm and th...

AI Technical Summary

Problems solved by technology

[0009] The present invention proposes a CNN speech enhancement algorithm based on voiceprint embedding, which is used to solve the problem of poor speech enhancement effect under low signal-to-noise ratio conditions, and promotes faster and better implementation of intelligent speech technology

Images