Sound wave signal noise reduction method and device, electronic equipment and computer storage medium
By performing short-time Fourier transform and complex convolutional neural network processing on construction acoustic signals, the problem of inaccurate acquisition of construction acoustic signals in the existing technology is solved, and accurate noise reduction and acquisition of signals is achieved, ensuring the reliability of pipeline safety inspection.
Patent Information
- Application Number
- CN202411665416.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-20
- Publication Date
- 2025-03-07
AI Technical Summary
The prior art cannot accurately collect the acoustic signals generated by normal construction, resulting in inaccurate collection of construction acoustic signals, affecting the safety inspection of pipelines.
By performing a short-time Fourier transform on the sound wave signal to be processed, the complex domain time-frequency signal is obtained, and the pre-trained complex convolutional neural network model is used to process it, the complex proportion masking value is calculated, the signal is enhanced, and the time-domain signal is finally obtained through the inverse short-time Fourier transform.
It realizes accurate noise reduction of construction acoustic signals, retains amplitude and phase information, improves the accuracy of acquisition of acoustic signals, and ensures the reliability of pipeline safety inspection.
Smart Images
Figure CN119580753A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of signal processing technology, and in particular to a method, device, electronic equipment and computer storage medium for reducing noise of an acoustic wave signal. Background Art
[0002] Fiber-optic distributed acoustic sensing (DAS) systems have gradually been used in safety monitoring of oil and gas pipelines. The main principle is to use fiber-optic sensing technology to sense the acoustic signals generated by third-party construction near the pipeline and identify them to determine whether there are safety hazards.
[0003] In the real environment, there are often various environmental interference sounds that will affect the recognition of sound wave signals. Therefore, the collected sound wave signals need to be denoised. Traditional noise reduction methods include filtering noise reduction, wavelet transform noise reduction, etc. These noise reduction methods have poor noise reduction effects in construction environments and other load environments. The sound waves generated by normal construction may be treated as noise, resulting in inaccurate collection of construction sound wave signals, affecting the safety inspection of pipelines.
[0004] This shows that the existing technology is unable to accurately collect the sound wave signals generated by normal construction. Summary of the invention
[0005] In view of this, it is necessary to provide a method, device, electronic device and computer storage medium for reducing noise of sound wave signals to solve the problem that the prior art cannot accurately collect the sound wave signals generated by normal construction.
[0006] In order to solve the above problems, the present invention provides a method for reducing noise of an acoustic wave signal, comprising: Acquiring a sound wave signal to be processed, wherein the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; Performing short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; Using a preset complex convolutional neural network model to process the complex domain time-frequency signal of the sound wave signal to be processed, to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; The complex domain time-frequency signal of the sound wave signal to be processed is enhanced by using the complex ratio masking value to obtain the complex domain time-frequency signal of the target sound wave signal; An inverse short-time Fourier transform is performed on the complex domain time-frequency signal of the target sound wave signal to obtain a time domain signal of the target sound wave signal.
[0007] In a possible implementation manner, performing a short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed includes: The sound wave signal to be processed is processed by using a short-time Fourier transform with a preset window length and a preset sliding window length, and the complex form of the short-time Fourier transform result is determined as the complex domain time-frequency signal of the sound wave signal to be processed.
[0008] In a possible implementation, the training process of the complex convolutional neural network model includes: The target sound wave signal and the noise sound wave signal are collected, and the target sound wave signal and the noise sound wave signal are edited, aligned, mixed and superimposed to obtain a noisy target sound wave signal; Calculating the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal, taking the noisy target sound wave signal as the input of the complex convolutional neural network model, and outputting the complex ratio masking calculation value of the noisy target sound wave signal and the target sound wave signal, and training the complex convolutional neural network model; Through iterative training, the difference between the complex ratio masking calculation value and the complex ratio masking target value is made to be within a preset error range.
[0009] In a possible implementation manner, the calculating the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal includes: The complex ratio masking calculation formula is used to calculate the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal. The complex ratio masking calculation formula is:
[0010] Where CRM is the complex ratio masking target value, is the complex imaginary unit, and represent the real and imaginary parts of the complex domain time-frequency signal of the noisy target sound wave signal obtained after short-time Fourier transform, and Respectively represent the real part and imaginary part of the complex domain time-frequency signal of the target sound wave signal obtained after short-time Fourier transform; The relationship between the target acoustic wave signal and the noisy target acoustic wave signal is:
[0011] Where E is the complex domain time-frequency signal of the target sound wave signal, is the complex domain time-frequency signal of the noisy target sound wave signal, is the complex scale masking value.
[0012] In a possible implementation, the using a preset complex convolutional neural network model to process the complex domain time-frequency signal includes: Using an encoder composed of multiple independent two-dimensional convolutional network modules, a normalization layer and an activation layer to encode the complex domain time-frequency signal of the sound wave signal to be processed, so as to obtain feature data of the complex domain time-frequency signal of the sound wave signal to be processed; A decoder composed of multiple independent two-dimensional convolutional network modules, a normalization layer and an activation layer is used to decode the characteristic data of the complex domain time-frequency signal of the processed sound wave signal to obtain the complex proportional masking values of the sound wave signal to be processed and the target sound wave signal.
[0013] In one possible implementation, the complex convolutional neural network model is further embedded with a long short-term memory network layer including 256 hidden units.
[0014] In a possible implementation manner, the preset window length is 64 milliseconds, and the preset sliding window length is 16 milliseconds.
[0015] The present invention also provides a device for reducing noise of an acoustic wave signal, comprising: A signal acquisition module, used to acquire a sound wave signal to be processed, wherein the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; A Fourier transform module, used for performing a short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; A calculation module, used for processing the complex domain time-frequency signal of the sound wave signal to be processed by using a preset complex convolutional neural network model to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; An enhancement module, configured to enhance the complex domain time-frequency signal of the sound wave signal to be processed by using the complex ratio masking value to obtain the complex domain time-frequency signal of the target sound wave signal; The inverse transform module is used to perform an inverse short-time Fourier transform on the complex domain time-frequency signal of the target sound wave signal to obtain the time domain signal of the target sound wave signal.
[0016] The present invention also provides an electronic device, comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the method for reducing noise of sound wave signals described in any of the above embodiments.
[0017] The present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the method for reducing noise of an acoustic wave signal described in any of the above embodiments.
[0018] The beneficial effects of the present invention are as follows: the method for reducing noise of the acoustic wave signal provided by the present invention obtains the complex domain time-frequency signal of the acoustic wave signal to be processed by performing a short-time Fourier transform on the acoustic wave signal to be processed, and then uses a pre-trained complex neural network model to identify the complex domain time-frequency signal of the acoustic wave signal to be processed, and obtains the complex ratio masking value of the acoustic wave signal to be processed and the target acoustic wave signal, and then enhances the complex domain time-frequency signal of the acoustic wave signal to be processed based on the complex ratio masking value to obtain the complex domain time-frequency signal of the target acoustic wave signal; performs an inverse short-time Fourier transform on the complex domain time-frequency signal of the target acoustic wave signal to obtain the time domain signal of the target acoustic wave signal. By reducing noise of the acoustic wave from the time-frequency domain after the short-time Fourier transform, the amplitude and phase information can be retained at the same time, and the complex convolutional neural network is used to perform convolution calculation on the complex domain in the data after the short-time Fourier transform, so that the target acoustic wave can be retained more accurately, and the accuracy of noise reduction of the acoustic wave signal can be ensured. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0020] Figure 1 A schematic diagram of a flow chart of a method for reducing noise of an acoustic wave signal provided by an embodiment of the present invention; Figure 2 A schematic diagram of a training process flow of a complex convolutional neural network model provided by an embodiment of the present invention; Figure 3 A schematic diagram of a flow chart of a method for calculating a complex ratio masking value provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of a complex convolutional neural network model provided by an embodiment of the present invention; Figure 5 A schematic diagram of the structure of a sound wave signal noise reduction device provided by an embodiment of the present invention; Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0021] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of the present invention and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.
[0022] In the description of the embodiments of the present invention, unless otherwise specified, "multiple" means two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone.
[0023] The descriptions of "first", "second", etc. involved in the embodiments of the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the technical features defined as "first" and "second" may explicitly or implicitly include at least one of the features.
[0024] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0025] A specific embodiment of the present invention, as Figure 1 As shown, a method for reducing noise of an acoustic wave signal is disclosed, comprising: S101, obtaining a sound wave signal to be processed, where the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; S102, performing short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; S103, using a preset complex convolutional neural network model to process the complex domain time-frequency signal of the sound wave signal to be processed, to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; S104, using a complex proportional masking value to enhance the complex domain time-frequency signal of the sound wave signal to be processed, to obtain a complex domain time-frequency signal of the target sound wave signal; S105, performing an inverse short-time Fourier transform on the complex domain time-frequency signal of the target sound wave signal to obtain a time domain signal of the target sound wave signal.
[0026] In an embodiment of the present invention, the sound wave signal to be processed refers to a mixed sound wave signal including a target sound wave signal and a noise sound wave signal. The sound wave signal to be processed can be obtained by collecting it through a sound wave sensor. For example, for safety monitoring in the field of oil and gas pipelines, an optical fiber distributed sound wave sensor can be used to collect sound wave signals near the pipeline. The sound wave signal is the sound wave signal to be processed, wherein the target sound wave signal in the sound wave signal is the sound wave signal generated during construction at a construction site near the pipeline, and the noise sound wave signal is other sound wave signals.
[0027] In an embodiment of the present invention, after obtaining the sound wave signal to be processed, a short-time Fourier transform is performed on the sound wave signal to be processed. Specifically, a signal in a certain time period of the sound wave signal to be processed can be selected for processing to obtain a complex domain time-frequency signal of the sound wave signal to be processed. Furthermore, a preset complex convolutional neural network model is used to process the complex domain time-frequency signal of the sound wave signal to be processed. The complex convolutional neural network model is a pre-trained deep learning model for calculating the complex proportional masking value of the sound wave signal to be processed and the target sound wave signal. The complex domain time-frequency signal of the target sound wave signal and the complex domain time-frequency signal of the sound wave signal to be processed can be converted through the complex proportional masking value. Furthermore, after determining the complex proportional masking values of the sound wave signal to be processed and the target sound wave signal, the complex domain time-frequency signal of the target sound wave signal is calculated based on the complex proportional masking value and the complex domain time-frequency signal of the sound wave signal to be processed, and then the complex domain time-frequency signal of the target sound wave signal is inversely short-time Fourier transformed to obtain a pure time domain signal of the target sound wave signal.
[0028] The method for reducing the noise of the acoustic wave signal provided by the present invention obtains the complex domain time-frequency signal of the acoustic wave signal to be processed by performing a short-time Fourier transform on the acoustic wave signal to be processed, and then uses a pre-trained complex neural network model to identify the complex domain time-frequency signal of the acoustic wave signal to be processed, and obtains the complex ratio masking value of the acoustic wave signal to be processed and the target acoustic wave signal, and then enhances the complex domain time-frequency signal of the acoustic wave signal to be processed based on the complex ratio masking value to obtain the complex domain time-frequency signal of the target acoustic wave signal; performs an inverse short-time Fourier transform on the complex domain time-frequency signal of the target acoustic wave signal to obtain the time domain signal of the target acoustic wave signal. By reducing the noise of the acoustic wave from the time-frequency domain after the short-time Fourier transform, the amplitude and phase information can be retained at the same time, and the complex convolutional neural network is used to perform convolution calculation on the complex domain in the data after the short-time Fourier transform, so that the target acoustic wave can be retained more accurately, and the accuracy of the noise reduction of the acoustic wave signal can be ensured.
[0029] As a possible implementation mode of the present invention, in this implementation mode, a short-time Fourier transform is performed on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed, including: The acoustic wave signal to be processed is processed by using a short-time Fourier transform with a preset window length and a preset sliding window length, and the complex form of the short-time Fourier transform result is determined as a complex domain time-frequency signal of the acoustic wave signal to be processed.
[0030] In an embodiment of the present invention, when a short-time Fourier transform is performed on a processed sound wave signal, a sound wave signal to be processed of a preset duration can be selected for processing, such as a sound wave signal to be processed of a length of 10 seconds, and a short-time Fourier transform of a preset window length and a preset sliding window length is used to process the sound wave signal to be processed, and the preset window length can be 64 milliseconds, and the preset sliding window length can be 16 milliseconds; specifically, the length of the sound wave signal to be processed, the preset window length, and the preset sliding window length can be determined according to actual conditions, and the present invention does not limit this. After the sound wave signal to be processed is short-time Fourier transform is performed, a complex signal including a real part and an imaginary part can be obtained, and the complex signal is determined as a complex domain time-frequency signal of the sound wave signal to be processed.
[0031] The embodiment of the present invention performs short-time Fourier transform on the sound wave signal to be processed, so that the amplitude and phase information can be retained at the same time when performing sound wave noise reduction, and the noise reduction effect is better.
[0032] As a possible implementation mode of the present invention, in this implementation mode, Figure 2 As shown in Figure 1, the training process of the complex convolutional neural network model includes: S201, collecting a target sound wave signal and a noise sound wave signal, editing, aligning, mixing and superimposing the target sound wave signal and the noise sound wave signal to obtain a noisy target sound wave signal; S202, calculating the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal, taking the noisy target sound wave signal as the input of the complex convolutional neural network model, and outputting the complex ratio masking calculation value of the noisy target sound wave signal and the target sound wave signal, and training the complex convolutional neural network model; S203, through iterative training, the difference between the complex ratio masking calculation value and the complex ratio masking target value is made to be within a preset error range.
[0033] In the implementation of the present invention, when training the complex convolutional neural network model, it is necessary to first construct a training set, and the data in the training set includes a target sound wave signal and a noise sound wave signal. Specifically, the corresponding target sound wave signal and noise sound wave signal can be collected for a specific application scenario. For example, for safety monitoring in the field of oil and gas pipelines, the construction sound wave signal of the construction site can be collected in a quiet environment and used as the target sound wave signal, and the noise sound wave signal when the construction site is not under construction can be collected, and then the target sound wave signal and the noise sound wave signal are edited and aligned, such as editing the corresponding 10-second target sound wave signal and noise sound wave signal, and then mixing and superimposing them to obtain a 10-second long noisy target sound wave signal.
[0034] Furthermore, the noisy sound wave signal and the target sound wave signal are short-time Fourier transformed to obtain a complex domain time-frequency signal of the noisy sound wave signal and a complex domain time-frequency signal of the target sound wave signal, respectively; then, based on the complex domain time-frequency signal of the noisy sound wave signal and the complex domain time-frequency signal of the target sound wave signal, a corresponding complex proportional masking target value is calculated; the complex domain time-frequency signal of the noisy target sound wave signal is used as an input of a complex convolutional neural network model; the complex proportional masking calculation value is used as an output to train the complex convolutional neural network model; and through iteration, the error between the complex proportional masking calculation value and the complex proportional masking target value is continuously made within a preset range, such as a mean square error between the complex proportional masking calculation value and the complex proportional masking target value is within a preset range.
[0035] The embodiment of the present invention trains the complex convolutional neural network model by collecting sound wave signals corresponding to the implementation environment to ensure the accuracy of the model.
[0036] As a possible implementation manner of the present invention, in this implementation manner, calculating the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal includes: The complex ratio masking calculation formula is used to calculate the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal. The complex ratio masking calculation formula is:
[0037] Where CRM is the complex ratio masking target value, is the complex imaginary unit, and represent the real and imaginary parts of the complex domain time-frequency signal of the noisy target sound wave signal obtained after short-time Fourier transform, and Respectively represent the real part and imaginary part of the complex domain time-frequency signal of the target sound wave signal obtained after short-time Fourier transform; The relationship between the target acoustic wave signal and the noisy target acoustic wave signal is:
[0038] Where E is the complex domain time-frequency signal of the target sound wave signal, is the complex domain time-frequency signal of the noisy target sound wave signal, is the complex scale masking value.
[0039] As a possible implementation mode of the present invention, Figure 3 As shown, a preset complex convolutional neural network model is used to process complex domain time-frequency signals, including: S301, using an encoder composed of multiple independent two-dimensional convolutional network modules, a normalization layer, and an activation layer to encode the complex domain time-frequency signal of the sound wave signal to be processed, to obtain feature data of the complex domain time-frequency signal of the sound wave signal to be processed; S302, using a decoder composed of multiple independent two-dimensional convolutional network modules, a normalization layer and an activation layer to decode the feature data of the complex domain time-frequency signal of the processed sound wave signal, and obtain the complex ratio masking value of the sound wave signal to be processed and the target sound wave signal.
[0040] In the embodiment of the present invention, Figure 4 As shown in the figure, the complex convolutional neural network is composed of multiple independent two-dimensional convolutional network modules, each module is responsible for processing a node in the complex multiplication operation, and each has a unique weight configuration. Four complex convolutional neural networks, a normalization layer and an activation layer constitute an encoder layer or a decoder layer. The real part and imaginary part of the complex domain time-frequency signal of the sound wave signal to be processed are used as the input of the complex convolutional neural network model. The complex domain time-frequency signal of the sound wave signal to be processed is encoded and decoded by the encoder and decoder to obtain the complex proportional masking values of the sound wave signal to be processed and the target sound wave signal.
[0041] Furthermore, the complex convolutional neural network model is also embedded with a long short-term memory network layer containing 256 hidden units to enhance the network's time series processing capabilities.
[0042] The acoustic signal denoising method provided by the present invention establishes a data set specifically for the acoustic wave signals around the pipeline collected by the optical fiber sensing technology, improves the accuracy of model recognition, performs denoising of the acoustic wave in the time-frequency domain after the short-time Fourier transform, and can simultaneously retain the amplitude and phase information. A deep learning model is constructed based on a complex convolutional neural network, and convolution calculation is also performed on the complex domain in the data after the short-time Fourier transform, which can more accurately retain the target acoustic wave and ensure the accuracy of the acoustic signal denoising.
[0043] In order to better implement the method for reducing noise of the acoustic wave signal in the embodiment of the present invention, based on the method for reducing noise of the acoustic wave signal, correspondingly, Figure 5 As shown, the embodiment of the present invention further provides a sound wave signal noise reduction device, and the sound wave signal noise reduction device 500 includes: The signal acquisition module 501 is used to acquire the sound wave signal to be processed, where the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; The Fourier transform module 502 is used to perform a short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; The calculation module 503 is used to process the complex domain time-frequency signal of the sound wave signal to be processed by using a preset complex convolutional neural network model to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; An enhancement module 504 is used to enhance the complex domain time-frequency signal of the sound wave signal to be processed by using a complex proportional masking value to obtain a complex domain time-frequency signal of the target sound wave signal; The inverse transform module 505 is used to perform an inverse short-time Fourier transform on the complex domain time-frequency signal of the target sound wave signal to obtain a time domain signal of the target sound wave signal.
[0044] The acoustic signal denoising device 500 provided in the above embodiment can implement the technical solution described in the above acoustic signal denoising method embodiment. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above acoustic signal denoising method embodiment, which will not be repeated here.
[0045] like Figure 6 As shown, the present invention also provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602 and a display 603. Figure 6 Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.
[0046] In some embodiments, the processor 601 may be a central processing unit (CPU), a microprocessor or other data processing chip, and is used to run program codes or process data stored in the memory 602, such as the acoustic signal noise reduction method of the present invention.
[0047] In some embodiments, the processor 601 may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor 601 may be local or remote. In some embodiments, the processor 601 may be implemented in a cloud platform. In some embodiments, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-cloud, etc., or any combination thereof.
[0048] In some embodiments, the memory 602 may be an internal storage unit of the electronic device 600, such as a hard disk or memory of the electronic device 600. In other embodiments, the memory 602 may also be an external storage device of the electronic device 600, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the electronic device 600.
[0049] Furthermore, the memory 602 may include both an internal storage unit of the electronic device 600 and an external storage device. The memory 602 is used to store application software installed in the electronic device 600 and various data.
[0050] In some embodiments, the display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 603 is used to display information of the electronic device 600 and to display a visual user interface. The components 601-603 of the electronic device 600 communicate with each other through a system bus.
[0051] In some embodiments, when the processor 601 executes the sound wave signal noise reduction program in the memory 602, the following steps may be implemented: Acquire a sound wave signal to be processed, where the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; Performing short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; A preset complex convolutional neural network model is used to process the complex domain time-frequency signal of the sound wave signal to be processed, so as to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; The complex domain time-frequency signal of the sound wave signal to be processed is enhanced by using a complex ratio masking value to obtain the complex domain time-frequency signal of the target sound wave signal; The complex domain time-frequency signal of the target sound wave signal is subjected to an inverse short-time Fourier transform to obtain the time domain signal of the target sound wave signal.
[0052] It should be understood that: when the processor 601 executes the sound wave signal noise reduction program in the memory 602, in addition to the above functions, other functions can also be implemented. For details, please refer to the description of the corresponding method embodiment above.
[0053] Furthermore, the embodiment of the present invention does not specifically limit the type of the electronic device 600 mentioned, and the electronic device 600 may be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of portable electronic devices include but are not limited to portable electronic devices equipped with IOS, Android, Microsoft or other operating systems. The above-mentioned portable electronic devices may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 600 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).
[0054] Correspondingly, an embodiment of the present invention also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the program or instructions are executed by a processor, the steps or functions in the sound wave signal noise reduction method provided in the above-mentioned method embodiments can be implemented.
[0055] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.
[0056] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.
Claims
1. A method for reducing noise of an acoustic wave signal, characterized in that: include: Acquiring a sound wave signal to be processed, wherein the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; Performing short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; Using a preset complex convolutional neural network model to process the complex domain time-frequency signal of the sound wave signal to be processed, to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; The complex domain time-frequency signal of the sound wave signal to be processed is enhanced by using the complex ratio masking value to obtain the complex domain time-frequency signal of the target sound wave signal; An inverse short-time Fourier transform is performed on the complex domain time-frequency signal of the target sound wave signal to obtain a time domain signal of the target sound wave signal.
2. The method for reducing noise of an acoustic wave signal according to claim 1, characterized in that: The step of performing a short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed includes: The sound wave signal to be processed is processed by using a short-time Fourier transform with a preset window length and a preset sliding window length, and the complex form of the short-time Fourier transform result is determined as the complex domain time-frequency signal of the sound wave signal to be processed.
3. The method for reducing noise of an acoustic wave signal according to claim 1, characterized in that: The training process of the complex convolutional neural network model includes: The target sound wave signal and the noise sound wave signal are collected, and the target sound wave signal and the noise sound wave signal are edited, aligned, mixed and superimposed to obtain a noisy target sound wave signal; Calculating the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal, taking the noisy target sound wave signal as the input of the complex convolutional neural network model, and outputting the complex ratio masking calculation value of the noisy target sound wave signal and the target sound wave signal, and training the complex convolutional neural network model; Through iterative training, the difference between the complex ratio masking calculation value and the complex ratio masking target value is made to be within a preset error range.
4. The method for reducing noise of an acoustic wave signal according to claim 3, characterized in that: The calculating of the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal comprises: The complex ratio masking calculation formula is used to calculate the complex ratio masking target value of the noisy target sound wave signal and the target sound wave signal. The complex ratio masking calculation formula is: Where CRM is the complex ratio masking target value, is the complex imaginary unit, and represent the real and imaginary parts of the complex domain time-frequency signal of the noisy target sound wave signal obtained after short-time Fourier transform, and Respectively represent the real part and imaginary part of the complex domain time-frequency signal of the target sound wave signal obtained after short-time Fourier transform; The relationship between the target acoustic wave signal and the noisy target acoustic wave signal is: Where E is the complex domain time-frequency signal of the target sound wave signal, is the complex domain time-frequency signal of the noisy target sound wave signal, is the complex scale masking value.
5. The method for reducing noise of sound wave signals according to claim 3, characterized in that: The using a preset complex convolutional neural network model to process the complex domain time-frequency signal includes: Using an encoder composed of multiple independent two-dimensional convolutional network modules, a normalization layer and an activation layer to encode the complex domain time-frequency signal of the sound wave signal to be processed, so as to obtain feature data of the complex domain time-frequency signal of the sound wave signal to be processed; A decoder composed of multiple independent two-dimensional convolutional network modules, a normalization layer and an activation layer is used to decode the characteristic data of the complex domain time-frequency signal of the processed sound wave signal to obtain the complex proportional masking values of the sound wave signal to be processed and the target sound wave signal.
6. The method for reducing noise of an acoustic wave signal according to claim 5, characterized in that: The complex convolutional neural network model is also embedded with a long short-term memory network layer containing 256 hidden units.
7. The method for reducing noise of an acoustic wave signal according to claim 2, characterized in that: The preset window length is 64 milliseconds, and the preset sliding window length is 16 milliseconds.
8. A device for reducing noise of acoustic wave signals, characterized in that: include: A signal acquisition module, used to acquire a sound wave signal to be processed, wherein the sound wave signal to be processed includes a noise sound wave signal and a target sound wave signal; A Fourier transform module, used for performing a short-time Fourier transform on the sound wave signal to be processed to obtain a complex domain time-frequency signal of the sound wave signal to be processed; A calculation module, used for processing the complex domain time-frequency signal of the sound wave signal to be processed by using a preset complex convolutional neural network model to obtain a complex proportional masking value of the sound wave signal to be processed and the target sound wave signal; An enhancement module, configured to enhance the complex domain time-frequency signal of the sound wave signal to be processed by using the complex ratio masking value to obtain the complex domain time-frequency signal of the target sound wave signal; The inverse transform module is used to perform an inverse short-time Fourier transform on the complex domain time-frequency signal of the target sound wave signal to obtain the time domain signal of the target sound wave signal.
9. An electronic device, characterized in that: comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the method for reducing noise of sound wave signals as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the sound wave signal noise reduction method described in any one of claims 1 to 7.