Speech signal processing method, device and storage medium

Abstract

The present invention relates to a speech signal processing method, device, equipment and storage medium, comprising: performing preliminary noise reduction processing on an input noisy speech signal to obtain an initial separated speech signal, respectively obtaining the power value of the noisy speech signal, the power value of the initial separated speech signal and the power value of the noise signal, and determining the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio after time domain smoothing, and then updating the noise reduction function corresponding to a noise reduction gainer according to the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio to obtain an updated noise reduction function, thereby realizing dynamic adjustment of the noise reduction gainer, finding a better gain value, and performing frequency domain smoothing on the updated noise reduction function to obtain a filter function corresponding to a noise reduction filter, and then performing inverse Fourier transform on the filter function and the noisy speech signal to obtain a target separated speech signal in the frequency domain, thereby obtaining a separated speech signal in the time domain, realizing noise reduction of the noisy signal under a better gain value, and improving noise reduction performance.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to a method, device, equipment and storage medium for processing a speech signal. Background Art

[0002] Speech noise reduction refers to the technology of extracting useful speech signals from the noise background and suppressing and reducing noise interference when the speech signal is interfered with or even drowned by various noises. Simply put, it is to extract the purest original speech from the noisy speech.

[0003] Normally, NN noise reduction can be used to reduce the noise of speech. However, NN noise reduction usually has two problems when the noise is strong: first, since NN noise reduction solves the time-frequency mask in a single frame, the continuity between frames is weak in principle, the data is manifested as a broken spectrum, and there is a "crackling" impact sound; second, the speech segment and the silent segment are processed unbalancedly, the noise suppression in the silent segment is better, but there is more residual noise in the speech segment, which makes the noise more abrupt in the hearing.

[0004] Therefore, the noise suppression effect in the prior art is relatively poor. Summary of the invention

[0005] The present invention provides a method, device, equipment and storage medium for processing a speech signal to solve the technical problem of poor noise suppression effect in the prior art.

[0006] The technical solution of the present invention to solve the above technical problems is as follows:

[0007] A method for processing a speech signal, comprising:

[0008] Performing preliminary noise reduction processing on the input noisy speech signal to obtain an initial separated speech signal, and respectively obtaining a power value of the noisy speech signal and a power value of the initial separated speech signal;

[0009] Determining a power value of the noise signal according to the power value of the noisy speech signal and the power value of the initial separated speech signal;

[0010] Respectively performing time domain smoothing on the power value of the initial separated speech signal, the power value of the noise signal, and the power value of the noisy speech signal to obtain smoothed power values ​​of the initial separated speech signal, the smoothed power value of the noise signal, and the power value of the noisy speech signal;

[0011] Determining a priori signal-to-noise ratio and a posteriori signal-to-noise ratio according to the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal;

[0012] According to the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio, updating the noise reduction function corresponding to the noise reduction gainer to obtain an updated noise reduction function;

[0013] Performing frequency domain smoothing on the updated denoising function to obtain a filter function corresponding to the denoising filter;

[0014] Obtaining a target separated speech signal in the frequency domain according to the filter function and the noisy speech signal;

[0015] Perform inverse Fourier transform on the target separated speech signal in the frequency domain to obtain a separated speech signal in the time domain.

[0016] Furthermore, in the above-mentioned method for processing speech signals, performing preliminary noise reduction processing on the input noisy speech signal to obtain an initial separated speech signal includes:

[0017] Performing preliminary noise reduction processing on the noisy speech signal to obtain a noise reduction signal;

[0018] Generating a time-frequency mask according to the noise reduction signal and the noisy speech signal;

[0019] The initial separated speech signal is generated according to the time-frequency mask and the noisy speech signal.

[0020] Furthermore, in the above-mentioned method for processing speech signals, the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal satisfy a first calculation formula;

[0021] The first calculation formula is:

[0022]

[0023] Among them, Es(t,f) represents the power value of the initial separated speech signal, S(t,f) represents the initial separated speech signal, En(t,f) represents the power value of the noise signal, Ex(t,f) represents the power value of the noisy speech signal, X(t,f) represents the noisy speech signal, and Y(t,f) represents the noise reduction signal.

[0024] Further, in the above-mentioned method for processing speech signals, the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal are respectively smoothed in the time domain to obtain the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal, including:

[0025] According to the second calculation formula, the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal are respectively smoothed in the time domain to obtain the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal;

[0026] The second calculation formula is:

[0027]

[0028] in, represents the smoothed power value of the initial separated speech signal at time t, represents the smoothed power value of the initial separated speech signal at time t-1, and α represents the smoothing coefficient of the initial separated speech signal; represents the smoothed power value of the noise signal at time t, represents the smoothed power value of the noise signal at time t-1, and β represents the smoothing coefficient of the noise signal; represents the smoothed power value of the noisy speech signal at time t, represents the smoothed power value of the noisy speech signal at time t-1, and δ represents the smoothing coefficient of the noisy speech signal.

[0029] Furthermore, in the above-mentioned method for processing speech signals, determining a priori signal-to-noise ratio and a posteriori signal-to-noise ratio according to the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal comprises:

[0030] Substituting the smoothed power value of the initial separated speech signal and the smoothed power value of the noise signal into a third calculation formula to obtain the priori signal-to-noise ratio; and substituting the smoothed power value of the power value of the noisy speech signal and the smoothed power value of the noise signal into a fourth calculation formula to obtain the posterior signal-to-noise ratio;

[0031] The third calculation formula is:

[0032]

[0033] ε(t,f) represents the prior signal-to-noise ratio;

[0034] The fourth calculation formula is:

[0035]

[0036] γ(t,f) represents the prior signal-to-noise ratio.

[0037] Furthermore, in the above-mentioned method for processing speech signals, updating the noise reduction function corresponding to the noise reduction gainer according to the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio to obtain an updated noise reduction function includes:

[0038] Substituting the a priori signal-to-noise ratio and the a posteriori signal-to-noise ratio into a fifth calculation formula to obtain the updated noise reduction function;

[0039] The fifth calculation formula is:

[0040]

[0041] Furthermore, in the above-mentioned method for processing speech signals, the updated noise reduction function is smoothed in the frequency domain to obtain a filter function corresponding to the noise reduction filter, including:

[0042] Substituting the updated noise reduction function into the sixth calculation formula, the filter function is obtained:

[0043] The sixth calculation formula is:

[0044]

[0045] Represents the filtering function, smoothMat represents the final normalized matrix of the Mel filter, MelMat represents the normalized transformation matrix of the Mel filter, and InvMelMat represents the normalized transposed matrix of the Mel filter.

[0046] Furthermore, in the above-mentioned method for processing speech signals, obtaining a target separated speech signal in the frequency domain according to the filter function and the noisy speech signal includes:

[0047] Substituting the filter function and the noisy speech signal into a seventh calculation formula to obtain a target separated speech signal in the frequency domain;

[0048] The seventh calculation formula is:

[0049]

[0050] Represents the target separated speech signal.

[0051] The present invention also provides a speech signal processing device, comprising: a processor and a memory;

[0052] The processor is used to execute the speech signal processing program stored in the memory to implement the speech signal processing method as described in any one of the above items.

[0053] The present invention also provides a storage medium, wherein the storage medium stores one or more programs, and the one or more programs can be executed to implement the method for processing a speech signal as described in any one of the above items.

[0054] The beneficial effects of the present invention are:

[0055] After performing preliminary noise reduction processing on the input noisy speech signal to obtain the initial separated speech signal, the power value of the noisy speech signal, the power value of the initial separated speech signal and the power value of the noise signal are obtained respectively, and after time domain smoothing, the priori signal-to-noise ratio and the posterior signal-to-noise ratio are determined, and then according to the priori signal-to-noise ratio and the posterior signal-to-noise ratio, the noise reduction function corresponding to the noise reduction gainer is updated to obtain the updated noise reduction function, thereby realizing dynamic adjustment of the noise reduction gainer, finding a better gain value, and performing frequency domain smoothing on the updated noise reduction function to obtain the filter function corresponding to the noise reduction filter, and after obtaining the target separated speech signal in the frequency domain according to the filter function and the noisy speech signal, an inverse Fourier transform is performed to obtain the separated speech signal in the time domain, thereby realizing noise reduction of the noisy signal at a better gain value and improving the noise reduction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flowchart of an embodiment of a method for processing a speech signal of the present invention;

[0057] Figure 2 A schematic diagram of the structure of an embodiment of a speech signal processing device of the present invention;

[0058] Figure 3 It is a structural schematic diagram of the speech signal processing device of the present invention. DETAILED DESCRIPTION

[0059] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0060] Figure 1 is a flow chart of an embodiment of a method for processing a speech signal of the present invention, as shown in FIG. Figure 1 As shown, the method for processing the speech signal of this embodiment may specifically include the following steps:

[0061] 100. Performing preliminary noise reduction processing on the input noisy speech signal to obtain an initial separated speech signal, and respectively obtaining a power value of the noisy speech signal and a power value of the initial separated speech signal;

[0062] In a specific implementation process, the noisy speech signal can be subjected to preliminary noise reduction processing to obtain a noise reduction signal; a time-frequency mask is generated based on the noise reduction signal and the noisy speech signal; and the initial separated speech signal is generated based on the time-frequency mask and the noisy speech signal.

[0063] Specifically, after the noisy speech signal is input, it is framed, windowed, and Fourier transformed to obtain the changed noisy speech signal X(t,f). The NN noise reduction algorithm can be used to perform preliminary noise reduction processing on the input noisy speech signal, and after framed, windowed, and Fourier transformed, the noise reduction signal Y(t,f) is obtained. The ratio of the noise reduction signal Y(t,f) to the noisy speech signal X(t,f) is used as the noisy speech signal Mask(t,f), and the product of the time-frequency mask and the noisy speech signal is used as the initial separated speech signal S(t,f).

[0064] Among them, S(t,f)=Mask(t,f)*X(t,f).

[0065] In a specific implementation process, after obtaining the initial separated speech signal, the power value of the noisy speech signal Es(t,f)=|S(t,f)| 2 and the power value of the initial separated speech signal Ex(t,f)=|X(t,f)| 2 .

[0066] 101. Determine a power value of a noise signal according to a power value of the noisy speech signal and a power value of the initial separated speech signal;

[0067] Specifically, the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal satisfy the first calculation formula (1);

[0068] The first calculation formula (1) is:

[0069]

[0070] Among them, Es(t,f) represents the power value of the initial separated speech signal, S(t,f) represents the initial separated speech signal, En(t,f) represents the power value of the noise signal, Ex(t,f) represents the power value of the noisy speech signal, X(t,f) represents the noisy speech signal, and Y(t,f) represents the noise reduction signal.

[0071] 102. Perform time domain smoothing on the power value of the initial separated speech signal, the power value of the noise signal, and the power value of the noisy speech signal, respectively, to obtain smoothed power values ​​of the initial separated speech signal, the smoothed power value of the noise signal, and the power value of the noisy speech signal;

[0072] Specifically, according to the second calculation formula (2), the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal can be respectively smoothed in the time domain to obtain the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the noisy speech signal.

[0073] The second calculation formula (2) is:

[0074]

[0075] in, represents the smoothed power value of the initial separated speech signal at time t, represents the smoothed power value of the initial separated speech signal at time t-1, and α represents the smoothing coefficient of the initial separated speech signal; represents the smoothed power value of the noise signal at time t, represents the smoothed power value of the noise signal at time t-1, and β represents the smoothing coefficient of the noise signal; represents the smoothed power value of the noisy speech signal at time t, represents the smoothed power value of the noisy speech signal at time t-1, and δ represents the smoothing coefficient of the noisy speech signal.

[0076] 103. Determine a priori signal-to-noise ratio and a posteriori signal-to-noise ratio according to the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal, and the smoothed power value of the power value of the noisy speech signal;

[0077] Specifically, the smoothed power value of the initial separated speech signal and the smoothed power value of the noise signal can be substituted into the third calculation formula (3) to obtain the priori signal-to-noise ratio; and the smoothed power value of the power value of the noisy speech signal and the smoothed power value of the noise signal can be substituted into the fourth calculation formula (4) to obtain the posterior signal-to-noise ratio;

[0078] The third calculation formula (3) is:

[0079]

[0080] ε(t,f) represents the prior signal-to-noise ratio;

[0081] The fourth calculation formula (4) is:

[0082]

[0083] γ(t,f) represents the prior signal-to-noise ratio.

[0084] 104. According to the a priori signal-to-noise ratio and the a posteriori signal-to-noise ratio, update the noise reduction function corresponding to the noise reduction gainer to obtain an updated noise reduction function;

[0085] Specifically, the priori signal-to-noise ratio and the posterior signal-to-noise ratio may be substituted into the fifth calculation formula (5) to obtain the updated noise reduction function;

[0086] The fifth calculation formula (5) is:

[0087]

[0088] 105. Perform frequency domain smoothing on the updated noise reduction function to obtain a filter function corresponding to the noise reduction filter;

[0089] Specifically, the updated noise reduction function is substituted into the sixth calculation formula (6) to obtain the filtering function:

[0090] The sixth calculation formula (6) is:

[0091]

[0092] Represents the filtering function, smoothMat represents the final normalized matrix of the Mel filter, MelMat represents the normalized transformation matrix of the Mel filter, and InvMelMat represents the normalized transposed matrix of the Mel filter.

[0093] 106. Obtain a target separated speech signal in the frequency domain according to the filter function and the noisy speech signal;

[0094] Specifically, the filter function and the noisy speech signal can be substituted into the seventh calculation formula (7) to obtain the target separated speech signal in the frequency domain;

[0095] The seventh calculation formula (7) is:

[0096]

[0097] Represents the target separated speech signal.

[0098] 107. Perform inverse Fourier transform on the target separated speech signal in the frequency domain to obtain a separated speech signal in the time domain.

[0099] The speech signal processing method of the present embodiment performs preliminary noise reduction processing on the input noisy speech signal to obtain the initial separated speech signal, and then obtains the power value of the noisy speech signal, the power value of the initial separated speech signal and the power value of the noise signal respectively, and after time domain smoothing, determines the priori signal-to-noise ratio and the posterior signal-to-noise ratio, and then updates the noise reduction function corresponding to the noise reduction gainer according to the priori signal-to-noise ratio and the posterior signal-to-noise ratio to obtain the updated noise reduction function, thereby realizing dynamic adjustment of the noise reduction gainer, finding a better gain value, and performing frequency domain smoothing on the updated noise reduction function to obtain the filter function corresponding to the noise reduction filter, and then according to the filter function and the noisy speech signal, after obtaining the target separated speech signal in the frequency domain, performing inverse Fourier transform to obtain the separated speech signal in the time domain, realizing noise reduction of the noisy signal at a better gain value, and improving the noise reduction performance.

[0100] It should be noted that the method of the embodiment of the present invention can be performed by a single device, such as a computer or a server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present invention, and the multiple devices will interact with each other to complete the method.

[0101] Figure 2 FIG. 1 is a schematic diagram of a structure of an embodiment of a speech signal processing device of the present invention. Figure 2 As shown, the speech signal processing device of this embodiment may include: a preliminary noise reduction module 20, a power determination module 21, a time domain smoothing module 22, a signal-to-noise ratio determination module 23, a frequency domain smoothing module 24, an update module 25, a post-processing module 26 and a transformation module 27.

[0102] A preliminary noise reduction module 20, used to perform preliminary noise reduction processing on the input noisy speech signal to obtain an initial separated speech signal;

[0103] Specifically, the noisy speech signal may be subjected to preliminary noise reduction processing to obtain a noise reduction signal; a time-frequency mask may be generated according to the noise reduction signal and the noisy speech signal; and the initial separated speech signal may be generated according to the time-frequency mask and the noisy speech signal.

[0104] A power determination module 21, used to obtain the power value of the noisy speech signal and the power value of the initial separated speech signal respectively, and determine the power value of the noise signal according to the power value of the noisy speech signal and the power value of the initial separated speech signal;

[0105] The power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal satisfy a first calculation formula;

[0106] The first calculation formula is:

[0107]

[0108] Among them, Es(t,f) represents the power value of the initial separated speech signal, S(t,f) represents the initial separated speech signal, En(t,f) represents the power value of the noise signal, Ex(t,f) represents the power value of the noisy speech signal, X(t,f) represents the noisy speech signal, and Y(t,f) represents the noise reduction signal.

[0109] A time domain smoothing module 22 is used to perform time domain smoothing on the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal, respectively, to obtain smoothed power values ​​of the initial separated speech signal, the smoothed power value of the noise signal and the power value of the noisy speech signal;

[0110] According to the second calculation formula, the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal are respectively smoothed in the time domain to obtain the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal;

[0111] The second calculation formula is:

[0112]

[0113] in, represents the smoothed power value of the initial separated speech signal at time t, represents the smoothed power value of the initial separated speech signal at time t-1, and α represents the smoothing coefficient of the initial separated speech signal; represents the smoothed power value of the noise signal at time t, represents the smoothed power value of the noise signal at time t-1, and β represents the smoothing coefficient of the noise signal; represents the smoothed power value of the noisy speech signal at time t, represents the smoothed power value of the noisy speech signal at time t-1, and δ represents the smoothing coefficient of the noisy speech signal.

[0114] A signal-to-noise ratio determination module 23, configured to determine a priori signal-to-noise ratio and a posteriori signal-to-noise ratio according to the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal;

[0115] Specifically, the smoothed power value of the initial separated speech signal and the smoothed power value of the noise signal are substituted into a third calculation formula to obtain the priori signal-to-noise ratio; and the smoothed power value of the power value of the noisy speech signal and the smoothed power value of the noise signal are substituted into a fourth calculation formula to obtain the posterior signal-to-noise ratio;

[0116] The third calculation formula is:

[0117]

[0118] ε(t,f) represents the prior signal-to-noise ratio;

[0119] The fourth calculation formula is:

[0120]

[0121] γ(t,f) represents the prior signal-to-noise ratio.

[0122] An updating module 25, configured to update the noise reduction function corresponding to the noise reduction gainer according to the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio to obtain an updated noise reduction function;

[0123] Specifically, the priori signal-to-noise ratio and the posterior signal-to-noise ratio are substituted into a fifth calculation formula to obtain the updated noise reduction function;

[0124] The fifth calculation formula is:

[0125]

[0126] A frequency domain smoothing module 24 is used to perform frequency domain smoothing on the updated denoising function to obtain a filter function corresponding to the denoising filter;

[0127] Specifically, the updated noise reduction function is substituted into the sixth calculation formula to obtain the filtering function:

[0128] The sixth calculation formula is:

[0129]

[0130] Represents the filtering function, smoothMat represents the final normalized matrix of the Mel filter, MelMat represents the normalized transformation matrix of the Mel filter, and InvMelMat represents the normalized transposed matrix of the Mel filter.

[0131] The post-processing module 26 obtains a target separated speech signal in the frequency domain according to the filter function and the noisy speech signal;

[0132] Specifically, the filter function and the noisy speech signal are substituted into the seventh calculation formula to obtain the target separated speech signal in the frequency domain;

[0133] The seventh calculation formula is:

[0134]

[0135] Represents the target separated speech signal.

[0136] The transformation module 27 is used to perform inverse Fourier transformation on the target separated speech signal in the frequency domain to obtain a separated speech signal in the time domain.

[0137] The device of the above embodiment is used to implement the corresponding method in the above embodiment. Its specific implementation scheme can refer to the method recorded in the above embodiment and the relevant description in the method embodiment, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0138] Figure 3 FIG. 1 is a structural diagram of a speech signal processing device of the present invention, as shown in FIG. Figure 3 As shown, the communication device of this embodiment may include: a processor 1010 and a memory 1020. Those skilled in the art will appreciate that the device may also include an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other in communication within the device via the bus 1050.

[0139] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0140] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0141] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0142] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0143] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0144] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0145] The present invention also provides a storage medium storing one or more programs, wherein the one or more programs can be executed to implement the method for processing the speech signal of the above embodiment.

[0146] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0147] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples. Within the scope of the present invention, the above embodiments or technical features in different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0148] In addition, to simplify the description and discussion, and in order not to obscure the present invention, known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, the devices may be shown in the form of block diagrams to avoid obscuring the present invention, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the present invention will be implemented (i.e., these details should be fully within the scope of understanding of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present invention, it will be apparent to those skilled in the art that the present invention may be implemented without these specific details or with variations in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0149] Although the invention has been described in conjunction with specific embodiments of the invention, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0150] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for processing a speech signal, characterized in that: include: Performing preliminary noise reduction processing on the input noisy speech signal to obtain an initial separated speech signal, and respectively obtaining a power value of the noisy speech signal and a power value of the initial separated speech signal; Determining a power value of the noise signal according to the power value of the noisy speech signal and the power value of the initial separated speech signal; Respectively performing time domain smoothing on the power value of the initial separated speech signal, the power value of the noise signal, and the power value of the noisy speech signal to obtain smoothed power values ​​of the initial separated speech signal, the smoothed power value of the noise signal, and the power value of the noisy speech signal; Determining a priori signal-to-noise ratio and a posteriori signal-to-noise ratio according to the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal; According to the a priori signal-to-noise ratio and the a posteriori signal-to-noise ratio, updating the noise reduction function corresponding to the noise reduction gainer to obtain an updated noise reduction function; Performing frequency domain smoothing on the updated denoising function to obtain a filter function corresponding to the denoising filter; Obtaining a target separated speech signal in the frequency domain according to the filter function and the noisy speech signal; Performing inverse Fourier transform on the target separated speech signal in the frequency domain to obtain a separated speech signal in the time domain; According to the a priori signal-to-noise ratio and the a posteriori signal-to-noise ratio, updating the noise reduction function corresponding to the noise reduction gainer to obtain an updated noise reduction function includes: Substituting the a priori signal-to-noise ratio and the a posteriori signal-to-noise ratio into a fifth calculation formula to obtain the updated noise reduction function; The fifth calculation formula is: Among them, ε(t,f) represents the prior signal-to-noise ratio, and γ(t,f) represents the posterior signal-to-noise ratio.

2. The method for processing speech signals according to claim 1, characterized in that: Perform preliminary noise reduction processing on the input noisy speech signal to obtain an initial separated speech signal, including: Performing preliminary noise reduction processing on the noisy speech signal to obtain a noise reduction signal; Generating a time-frequency mask according to the noise reduction signal and the noisy speech signal; The initial separated speech signal is generated according to the time-frequency mask and the noisy speech signal.

3. The method for processing speech signals according to claim 2, characterized in that: The power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal satisfy a first calculation formula; The first calculation formula is: Among them, Es(t,f) represents the power value of the initial separated speech signal, S(t,f) represents the initial separated speech signal, En(t,f) represents the power value of the noise signal, Ex(t,f) represents the power value of the noisy speech signal, X(t,f) represents the noisy speech signal, and Y(t,f) represents the noise reduction signal.

4. The method for processing speech signals according to claim 3, characterized in that: The respectively performing time domain smoothing on the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal to obtain the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal comprises: According to the second calculation formula, the power value of the initial separated speech signal, the power value of the noise signal and the power value of the noisy speech signal are respectively smoothed in the time domain to obtain the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal; The second calculation formula is: in, represents the smoothed power value of the initial separated speech signal at time t, represents the smoothed power value of the initial separated speech signal at time t-1, and α represents the smoothing coefficient of the initial separated speech signal; represents the smoothed power value of the noise signal at time t, represents the smoothed power value of the noise signal at time t-1, and β represents the smoothing coefficient of the noise signal; represents the smoothed power value of the noisy speech signal at time t, represents the smoothed power value of the noisy speech signal at time t-1, and δ represents the smoothing coefficient of the noisy speech signal.

5. The method for processing speech signals according to claim 4, characterized in that: Determining a priori signal-to-noise ratio and a posteriori signal-to-noise ratio according to the smoothed power value of the initial separated speech signal, the smoothed power value of the noise signal and the smoothed power value of the power value of the noisy speech signal, comprising: Substituting the smoothed power value of the initial separated speech signal and the smoothed power value of the noise signal into a third calculation formula to obtain the priori signal-to-noise ratio; and substituting the smoothed power value of the power value of the noisy speech signal and the smoothed power value of the noise signal into a fourth calculation formula to obtain the posterior signal-to-noise ratio; The third calculation formula is: The fourth calculation formula is:

6. The method for processing speech signals according to claim 1, characterized in that: The updated denoising function is smoothed in the frequency domain to obtain a filter function corresponding to the denoising filter, including: Substituting the updated noise reduction function into the sixth calculation formula, the filter function is obtained: The sixth calculation formula is: Represents the filtering function, smoothMat represents the final normalized matrix of the Mel filter, MelMat represents the normalized transformation matrix of the Mel filter, and InvMelMat represents the normalized transposed matrix of the Mel filter.

7. The method for processing speech signals according to claim 6, characterized in that: Obtaining a target separated speech signal in the frequency domain according to the filter function and the noisy speech signal, including: Substituting the filter function and the noisy speech signal into a seventh calculation formula to obtain a target separated speech signal in the frequency domain; The seventh calculation formula is: Represents the target separated speech signal.

8. A speech signal processing device, characterized in that: include: Processor and memory; The processor is used to execute the speech signal processing program stored in the memory to implement the speech signal processing method according to any one of claims 1 to 7.

9. A storage medium, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed to implement the method for processing a speech signal according to any one of claims 1 to 7.

Images