Voice noise reduction method and device and electronic equipment
By estimating and denoising the amplitude spectrum of the voice signal, the problem of poor results in complex noise environments is solved, and the degree of automation and noise reduction effect is improved.
Patent Information
- Application Number
- CN202411675242.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-21
- Publication Date
- 2025-02-21
AI Technical Summary
Existing voice noise reduction technology is not effective in dealing with nonlinear noise and complex noise environments. It requires a pre-understanding of the noise type and signal-to-noise ratio, and has a low degree of automation.
By estimating the amplitude spectrum in the speech signal, determining the amplitude spectrum mask and denoising the noise based on the mask, amplitude spectrum and phase spectrum, the advance understanding of the noise type and signal-to-noise ratio is avoided and the Requirement for engineers to manually optimize parameters.
The automation degree and noise reduction effect of the speech noise reduction method are improved, and can provide more satisfactory noise reduction effect in non-steady noise environments.
Smart Images
Figure CN119495313A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of speech noise reduction, and in particular to a speech noise reduction method, device and electronic equipment. Background Art
[0002] Existing speech noise reduction technologies generally use statistical models to eliminate noise from speech received by electronic devices, such as by using Gaussian mixture models or hidden Markov models to model speech signals and noise, thereby achieving noise reduction effects. However, the above methods are not effective when dealing with nonlinear noise and complex noise environments. Furthermore, the existing technology requires prior knowledge of the noise type and signal-to-noise ratio, and suppresses the noise on this basis, so the existing technology performs poorly in non-steady-state noise environments. Furthermore, in the presence of multiple noise types and time-varying noise, traditional methods often cannot provide satisfactory noise reduction effects. In addition, some traditional methods require engineers to manually optimize parameters, which requires professional knowledge and experience, resulting in a low degree of automation in existing speech noise reduction methods.
[0003] In general, the existing speech noise reduction methods have a low degree of automation and poor noise reduction effect. Summary of the invention
[0004] The purpose of the present invention is to provide a speech noise reduction method, device and electronic device to alleviate the technical problems of low automation and poor noise reduction effect of existing speech noise reduction methods, so as to improve the automation and noise reduction effect of the speech noise reduction method.
[0005] In the first aspect, an embodiment of the present invention provides a speech denoising method, comprising: obtaining original speech data; preprocessing the original speech data to obtain preprocessed audio data; performing Fourier transform on the preprocessed audio data to determine an amplitude spectrum and a phase spectrum corresponding to the preprocessed audio data; estimating noise based on the amplitude spectrum to determine an amplitude spectrum mask; and determining the denoised speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum.
[0006] In a preferred embodiment, the step of performing Fourier transform on the preprocessed audio data to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data includes: dividing the preprocessed audio data into short-time speech frames based on a preset number of frames; performing short-time Fourier transform on the short-time speech frames to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data.
[0007] In a preferred implementation, the value of the transformation points of the short-time Fourier transform is 512.
[0008] In a preferred embodiment, the step of performing noise estimation based on the above-mentioned amplitude spectrum and determining the amplitude spectrum mask includes: extracting the non-DC part in the above-mentioned amplitude spectrum to obtain complex spectrum data; performing noise estimation on the above-mentioned complex spectrum data through a preset complex operation model to determine the amplitude spectrum mask; wherein the above-mentioned preset complex operation model includes: an encoder in N layers of complex form, a recurrent neural network in preset M layers of complex form and a decoder in preset N layers of complex form.
[0009] In a preferred implementation, the value of N is 4; the value of M is 2.
[0010] In a preferred embodiment, the number of input channels of the above-mentioned recurrent neural network is equal to the number of input channels of the above-mentioned encoder; and the number of hidden nodes of the above-mentioned recurrent neural network is 128.
[0011] In a preferred embodiment, the above-mentioned recurrent neural network includes: a first amplitude spectrum recurrent neural network and a second amplitude spectrum recurrent neural network; the input nodes of the above-mentioned first amplitude spectrum recurrent neural network and the above-mentioned second amplitude spectrum recurrent neural network are equal; the number of hidden nodes of the above-mentioned first amplitude spectrum recurrent neural network and the above-mentioned second amplitude spectrum recurrent neural network are equal.
[0012] In a preferred embodiment, the step of determining the denoised speech data corresponding to the above-mentioned original speech data based on the above-mentioned amplitude spectrum mask, the above-mentioned amplitude spectrum and the above-mentioned phase spectrum includes: performing short-time Fourier inverse transform on the above-mentioned amplitude spectrum mask, the above-mentioned amplitude spectrum and the above-mentioned phase spectrum to obtain the denoised speech data corresponding to the above-mentioned original speech data.
[0013] In the second aspect, an embodiment of the present invention also provides a speech noise reduction device, including: a data acquisition module, used to acquire original speech data; a preprocessing module, used to preprocess the above original speech data to obtain preprocessed audio data; a Fourier transform module, used to perform Fourier transform on the above preprocessed audio data, and determine the amplitude spectrum and phase spectrum corresponding to the above preprocessed audio data; an amplitude spectrum mask determination module, used to perform noise estimation based on the above amplitude spectrum, and determine the amplitude spectrum mask; a noise reduction module, used to determine the noise-reduced speech data corresponding to the above original speech data based on the above amplitude spectrum mask, the above amplitude spectrum and the above phase spectrum.
[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the speech noise reduction method.
[0015] The embodiments of the present invention have the following beneficial technical effects:
[0016] The embodiments of the present invention provide a speech denoising method, device and electronic device, comprising: obtaining original speech data; preprocessing the original speech data to obtain preprocessed audio data; performing Fourier transform on the preprocessed audio data to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data; performing noise estimation based on the amplitude spectrum to determine the amplitude spectrum mask; and determining the denoised speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum. The method estimates the noise of the amplitude spectrum in the speech signal and determines the amplitude spectrum mask, so as to perform denoising on the original speech signal according to the amplitude spectrum mask, thereby eliminating the need to know the noise type and signal-to-noise ratio in advance. Moreover, the process does not require engineers to manually optimize parameters, thereby improving the degree of automation and denoising effect of the speech denoising method.
[0017] Other features and advantages disclosed in this embodiment will be described in the subsequent description, or some features and advantages can be inferred or determined without doubt from the description, or can be learned by implementing the above-mentioned technology of the present disclosure.
[0018] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0020] Figure 1 A flow chart of a speech noise reduction method provided by an embodiment of the present invention;
[0021] Figure 2 A flowchart of another speech noise reduction method provided by an embodiment of the present invention;
[0022] Figure 3 A schematic diagram of the structure of a speech noise reduction device provided by an embodiment of the present invention;
[0023] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention.
[0024] Icons: 31 - data acquisition module; 32 - preprocessing module; 33 - Fourier transform module; 34 - amplitude spectrum mask determination module; 35 - noise reduction module; 41 - processor; 42 - memory; 43 - bus; 44 - communication interface. DETAILED DESCRIPTION
[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0026] At present, the existing speech noise reduction technology generally uses a statistical model to eliminate noise from the speech received by the electronic device, for example, by using a Gaussian mixture model or a hidden Markov model to model the speech signal and noise, so as to achieve a noise reduction effect. However, the above method is not effective when dealing with nonlinear noise and complex noise environments. Furthermore, the existing technology requires a prior understanding of the noise type and signal-to-noise ratio, and suppresses the noise on this basis, so the existing technology performs poorly in a non-steady-state noise environment. Furthermore, in the presence of multiple noise types and time-varying noise, traditional methods often cannot provide satisfactory noise reduction effects. In addition, some traditional methods require engineers to manually optimize parameters, which requires professional knowledge and experience, resulting in a low degree of automation in existing speech noise reduction methods.
[0027] Based on this, an embodiment of the present invention provides a speech noise reduction method, device and electronic device. The method estimates the noise of the amplitude spectrum in the speech signal and determines the amplitude spectrum mask to perform noise reduction processing on the original speech signal according to the amplitude spectrum mask, so that there is no need to know the noise type and signal-to-noise ratio in advance. In addition, the process does not require engineers to manually optimize parameters, thereby improving the degree of automation and noise reduction effect of the speech noise reduction method. To facilitate understanding of the embodiments of the present invention, a speech noise reduction method disclosed in an embodiment of the present invention is first introduced in detail.
[0028] Example 1
[0029] Figure 1 A flow chart of a speech noise reduction method provided by an embodiment of the present invention.
[0030] Depend on Figure 1 As can be seen, the method comprises the following steps:
[0031] Step S101: Acquire original voice data.
[0032] In this embodiment, the original voice data is obtained through a microphone or a recording device.
[0033] Step S102: pre-processing the original speech data to obtain pre-processed audio data.
[0034] In actual operation, the original voice data is subjected to DC offset removal, sampling rate adjustment and audio normalization processing to obtain pre-processed audio data.
[0035] Here, the above preprocessing operation is performed on the original speech data to ensure that the signal is within an appropriate range.
[0036] Step S103: performing Fourier transform on the preprocessed audio data to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data.
[0037] Step S104: performing noise estimation according to the amplitude spectrum and determining an amplitude spectrum mask.
[0038] Step S105: Determine the noise-reduced speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum.
[0039] An embodiment of the present invention provides a speech denoising method, comprising: obtaining original speech data; preprocessing the original speech data to obtain preprocessed audio data; performing Fourier transform on the preprocessed audio data to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data; performing noise estimation based on the amplitude spectrum to determine the amplitude spectrum mask; and determining the denoised speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum. The method estimates the noise on the amplitude spectrum in the speech signal and determines the amplitude spectrum mask, so as to perform denoising on the original speech signal according to the amplitude spectrum mask, thereby eliminating the need to know the noise type and signal-to-noise ratio in advance. Moreover, the process does not require engineers to manually optimize parameters, thereby improving the degree of automation and denoising effect of the speech denoising method.
[0040] Example 2
[0041] Based on Example 1, Figure 2 A flowchart of another speech noise reduction method provided by an embodiment of the present invention.
[0042] Depend on Figure 2 As can be seen, the method includes:
[0043] Step S201: Acquire original voice data.
[0044] Step S202: pre-process the original voice data to obtain pre-processed audio data.
[0045] Step S203: Divide the pre-processed audio data into short-time speech frames based on a preset number of frames.
[0046] Step S204: performing short-time Fourier transform on the short-time speech frame to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data.
[0047] In this embodiment, the value of the transformation points of the short-time Fourier transform is 512.
[0048] Step S205: perform noise estimation based on the amplitude spectrum and determine the amplitude spectrum mask.
[0049] In this embodiment, step S205 includes the following steps A1-A2:
[0050] Step A1: extract the non-DC part in the amplitude spectrum to obtain complex spectrum data.
[0051] Step A2: Perform noise estimation on the complex spectrum data through a preset complex operation model to determine the amplitude spectrum mask; wherein the preset complex operation model includes: an N-layer complex encoder, a preset M-layer complex recurrent neural network and a preset N-layer complex decoder.
[0052] Furthermore, the number of input channels of the above-mentioned recurrent neural network is equal to the number of input channels of the above-mentioned encoder; the number of hidden nodes of the above-mentioned recurrent neural network is 128.
[0053] Furthermore, the above recurrent neural network includes: a first amplitude spectrum recurrent neural network and a second amplitude spectrum recurrent neural network;
[0054] The input nodes of the first amplitude spectrum recurrent neural network and the second amplitude spectrum recurrent neural network are equal; the number of hidden nodes of the first amplitude spectrum recurrent neural network and the second amplitude spectrum recurrent neural network are equal.
[0055] In actual operation, the number of output channels of Encoder1 to Encoder4 of the amplitude spectrum is preferably 16, 32, 64 and 64. i , whose input channel number is defined as the Decoder of the amplitude spectrum i+1 The number of output channels and the magnitude spectrum of the Encoder i+1 The sum of the number of output channels of . In the present invention, the number of output channels of amplitude spectrum Decoder1 to Decoder4 is preferably 2, 16, 32 and 64 respectively. The number of input nodes of RNN1 of amplitude spectrum is the output size of Encodern, and the number of hidden nodes is P, where P is 128. Furthermore, the last layer of amplitude spectrum RNN includes a fully connected layer, in which the number of output nodes of the fully connected layer is consistent with the number of input nodes of the first layer of amplitude spectrum RNN.
[0056] Furthermore, the method introduces amplitude masking to reduce the complexity of model learning, which is beneficial to model training and realizes real-time noise reduction and suppression of speech.
[0057] Step S206: Determine the noise-reduced speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum.
[0058] In this embodiment, the step S206 includes: performing short-time Fourier inverse transformation on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum to obtain the noise-reduced speech data corresponding to the original speech data.
[0059] The embodiment of the present invention provides a method for speech denoising, comprising: obtaining original speech data; preprocessing the original speech data to obtain preprocessed audio data; dividing the preprocessed audio data into speech short-time frames based on a preset number of frames; performing short-time Fourier transform on the speech short-time frames to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data; performing noise estimation based on the amplitude spectrum to determine the amplitude spectrum mask; and determining the denoised speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum. The method determines the amplitude spectrum by performing short-time Fourier transform on the speech short-time frames corresponding to the original speech data, and then determines the amplitude spectrum mask by performing noise estimation on the amplitude spectrum, so as to perform denoising on the original speech signal according to the amplitude spectrum mask, thereby eliminating the need to know the noise type and signal-to-noise ratio in advance. Moreover, the process does not require engineers to manually optimize parameters, thereby improving the degree of automation and denoising effect of the speech denoising method.
[0060] Example 3
[0061] In this embodiment, Figure 3 A schematic diagram of the structure of a speech noise reduction device provided in an embodiment of the present invention.
[0062] Depend on Figure 3 As can be seen, the device includes:
[0063] The data acquisition module 31 is used to acquire original voice data.
[0064] In this embodiment, the data acquisition module 31 includes: a microphone and an ADC hardware chip; the original voice data is a digital signal.
[0065] The preprocessing module 32 preprocesses the original voice data to obtain preprocessed audio data.
[0066] The Fourier transform module 33 performs Fourier transform on the pre-processed audio data to determine the amplitude spectrum and phase spectrum corresponding to the pre-processed audio data.
[0067] The amplitude spectrum mask determination module 34 is used to perform noise estimation according to the amplitude spectrum and determine the amplitude spectrum mask.
[0068] The noise reduction module 35 is used to determine the noise-reduced speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum.
[0069] The data acquisition module 31, the preprocessing module 32, the Fourier transform module 33, the amplitude spectrum mask determination module 34 and the noise reduction module 35 are sequentially connected.
[0070] In one embodiment, the Fourier transform module 33 is further used to divide the preprocessed audio data into short-time speech frames based on a preset number of frames; perform short-time Fourier transform on the short-time speech frames to determine the amplitude spectrum and phase spectrum corresponding to the preprocessed audio data.
[0071] In one embodiment, the amplitude spectrum mask determination module 34 is also used to extract the non-DC part of the amplitude spectrum to obtain complex spectrum data; the noise of the complex spectrum data is estimated by a preset complex operation model to determine the amplitude spectrum mask; wherein the preset complex operation model includes: an N-layer complex form encoder, a preset M-layer complex form recurrent neural network and a preset N-layer complex form decoder.
[0072] Furthermore, the noise reduction module 35 is also used to perform short-time Fourier inverse transformation on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum to obtain the noise-reduced speech data corresponding to the original speech data.
[0073] The embodiment of the present invention further provides an electronic device, such as Figure 4 As shown, it is a schematic diagram of the structure of the electronic device, wherein the electronic device includes a processor 41 and a memory 42, the memory 42 stores machine executable instructions that can be executed by the processor 41, and the processor 41 executes the machine executable instructions to implement the above-mentioned speech noise reduction method.
[0074] exist Figure 4 In the illustrated embodiment, the electronic device further includes a bus 43 and a communication interface 44 , wherein the processor 41 , the communication interface 44 and the memory 42 are connected via the bus.
[0075] The memory 42 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 44 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used. The bus may be an ISA bus, a PCI bus, or an EISA bus, etc. The above-mentioned bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0076] The processor 41 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 41. The above processor 41 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor 41 reads the information in the memory 42 and completes the steps of the speech noise reduction method of the above embodiment in combination with its hardware.
[0077] An embodiment of the present invention further provides a machine-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the above-mentioned speech noise reduction method. The specific implementation can be found in the aforementioned method embodiment, which will not be repeated here.
[0078] The computer program product of the speech noise reduction method, speech noise reduction device and electronic device provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the above program code can be used to execute the above speech noise reduction method in the previous method embodiment. The specific implementation can be referred to the method embodiment, which will not be repeated here.
[0079] Unless otherwise specifically stated, the relative steps, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the present invention.
[0080] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that can be executed by a processor. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0081] Finally, it should be noted that the above-mentioned embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-mentioned embodiments, ordinary technicians in this field should understand that any technician familiar with this technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed by the present invention, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.
Claims
1. A speech noise reduction method, characterized in that: include: Get the original voice data; Preprocessing the original voice data to obtain preprocessed audio data; Performing Fourier transform on the preprocessed audio data to determine an amplitude spectrum and a phase spectrum corresponding to the preprocessed audio data; Perform noise estimation based on the amplitude spectrum to determine an amplitude spectrum mask; Based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum, the noise-reduced speech data corresponding to the original speech data is determined.
2. The speech noise reduction method according to claim 1, characterized in that: The step of performing Fourier transform on the preprocessed audio data to determine an amplitude spectrum and a phase spectrum corresponding to the preprocessed audio data comprises: Dividing the preprocessed audio data into speech short-time frames based on a preset number of frames; Performing a short-time Fourier transform on the short-time speech frame to determine an amplitude spectrum and a phase spectrum corresponding to the preprocessed audio data.
3. The speech noise reduction method according to claim 2, characterized in that: The value of the transformation point number of the short-time Fourier transform is 512.
4. The speech noise reduction method according to claim 1, characterized in that: The step of performing noise estimation according to the amplitude spectrum and determining an amplitude spectrum mask comprises: Extracting the non-DC part in the amplitude spectrum to obtain complex spectrum data; Noise estimation is performed on the complex spectrum data through a preset complex operation model to determine the amplitude spectrum mask; wherein the preset complex operation model includes: an N-layer complex form encoder, a preset M-layer complex form recurrent neural network and a preset N-layer complex form decoder.
5. The speech noise reduction method according to claim 4, characterized in that: The value of N is 4; the value of M is 2.
6. The speech noise reduction method according to claim 4, characterized in that: The number of input channels of the recurrent neural network is equal to the number of input channels of the encoder; the number of hidden nodes of the recurrent neural network is 128.
7. The speech noise reduction method according to claim 6, characterized in that: The recurrent neural network comprises: a first amplitude spectrum recurrent neural network and a second amplitude spectrum recurrent neural network; The input nodes of the first amplitude spectrum recurrent neural network and the second amplitude spectrum recurrent neural network are equal; the number of hidden nodes of the first amplitude spectrum recurrent neural network and the second amplitude spectrum recurrent neural network are equal.
8. The speech noise reduction method according to claim 1, characterized in that: The step of determining the noise-reduced speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum comprises: Performing short-time inverse Fourier transform on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum to obtain noise-reduced speech data corresponding to the original speech data.
9. A speech noise reduction device, characterized in that: include: A data acquisition module, used to acquire original voice data; A preprocessing module preprocesses the original voice data to obtain preprocessed audio data; A Fourier transform module, performing a Fourier transform on the pre-processed audio data to determine an amplitude spectrum and a phase spectrum corresponding to the pre-processed audio data; An amplitude spectrum mask determination module, used for performing noise estimation based on the amplitude spectrum and determining an amplitude spectrum mask; The denoising module is used to determine the denoised speech data corresponding to the original speech data based on the amplitude spectrum mask, the amplitude spectrum and the phase spectrum.
10. An electronic device, characterized in that: The electronic device includes a processor and a memory, the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the speech noise reduction method according to any one of claims 1 to 8.