Unlock instant, AI-driven research and patent intelligence for your innovation.

Speech denoising method, device, equipment and medium based on improved gan network

A speech denoising and network technology, applied in speech analysis, instruments, etc., can solve problems such as unsatisfactory denoising results and inability to effectively denoise

Active Publication Date: 2019-11-26
BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a speech denoising method, device, equipment and medium based on an improved GAN network, which solves the problem that the method for denoising speech in the prior art can only remove the noise of simple distribution, and for complex distribution The noise signal cannot be effectively denoised, and valuable speech may be removed, resulting in unsatisfactory denoising results.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Speech denoising method, device, equipment and medium based on improved gan network
  • Speech denoising method, device, equipment and medium based on improved gan network
  • Speech denoising method, device, equipment and medium based on improved gan network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0027] figure 1 The flow chart of the speech denoising method based on the improved GAN network provided by Embodiment 1 of the present application, such as figure 1 As shown, the executor of the embodiment of the present application is the speech denoising device based on the improved GAN network, and the speech denoising device based on the improved GAN network can be integrated in the terminal device. The terminal device can be a smart phone, a vehicle terminal, a smart voice device, etc., and the smart voice device can be a smart voice speaker, a smart voice TV, a smart voice refrigerator, etc. Then, the speech denoising method based on the improved GAN network provided in this embodiment includes the following steps.

[0028] Step 201, acquire voice data to be processed.

[0029] Specifically, in this embodiment, the voice data to be processed may be user voice data, such as instruction voice data issued by the user. The speech data to be processed has noise, and the n...

Embodiment 2

[0047] image 3 The flow chart of the voice denoising method based on the improved GAN network provided in Embodiment 2 of the present application, as image 3 As shown, the speech denoising method based on the improved GAN network provided in this embodiment is based on the speech denoising method based on the improved GAN network provided in Embodiment 1 of the present application, further refinement of step 202, and also It includes the step of training and testing the GAN network until the GAN network converges, so as to obtain the step of improving the GAN network and the step of performing speech recognition on the denoised speech data. Then, the speech denoising method based on the improved GAN network provided in this embodiment includes the following steps.

[0048] Step 301, train and test the GAN network until the GAN network converges to obtain an improved GAN network.

[0049] Further, in this embodiment, the generator of the GAN network and the discriminator of...

Embodiment 3

[0103] Figure 5 A schematic structural diagram of a speech denoising device based on an improved GAN network provided in Embodiment 3 of the present application, as shown in Figure 5 As shown, the speech denoising device based on the improved GAN network provided in this embodiment includes: a data acquisition module 51 , a feature extraction module 52 , a processing value calculation module 53 , a speech denoising module 54 , and a denoising data determination module 55 .

[0104] Wherein, the data obtaining module 51 is used for obtaining the voice data to be processed. The feature extraction module 52 is configured to perform feature extraction on the speech data to be processed to form feature data of the speech to be processed. The processing value calculation module 53 is used to calculate the mean variance normalized processing value of the feature data of the speech to be processed. Speech denoising module 54, for inputting the mean variance normalized processing v...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The embodiment of the invention provides a voice denoising method, device, equipment and medium based on an improved GAN. The method includes: acquiring to-be-processed voice data; performing featureextraction of the to-be-processed voice data to form feature data of the to-be-processed voice; calculating a mean variance normalized processing value of the feature data of the to-be-processed voice; inputting the mean variance normalized processing value of the feature data of the to-be-processed voice into a generator of the improved GAN, and outputting an ideal mask value of the denoised voice feature data corresponding to the to-be-processed voice data; determining the denoised voice data of the to-be-processed voice data according to the ideal mask value of the denoised voice feature data, wherein the ideal mask value of the denoised voice feature data corresponding to the to-be-processed voice data is the ratio of the denoised voice feature data corresponding to the to-be-processedvoice data to the to-be-processed voice feature data. The invention can also have an obvious denoising effect for a complex distributed noise signal, and effectively improve the denoising effect.

Description

technical field [0001] The embodiment of the present application relates to the technical field of speech enhancement, and in particular to a speech denoising method, device, device and medium based on an improved GAN network. Background technique [0002] Speech enhancement refers to the technology of extracting useful speech signals from the noise background to suppress and reduce noise interference when the speech signal is disturbed or even submerged by various noises. The most important point of speech enhancement is to perform noise filtering on noisy speech to improve the clarity of sentences and the accuracy of speech recognition. [0003] In the prior art, there are mainly two methods for denoising speech: the traditional method of applying signal processing, and the advanced method of using deep learning models. Existing advanced methods using deep learning models generally use deep neural network models, long-term short-term memory network models, and convolution...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G10L21/0208G10L21/02
CPCG10L21/02G10L21/0208
Inventor 成学军
Owner BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD