Speech signal single-channel dereverberation method and device, terminal, and readable storage medium

Abstract

This invention discloses a method, apparatus, terminal, and readable storage medium for single-channel de-reverberation of speech signals. The method includes: acquiring a single-channel time-domain signal and performing frame segmentation, windowing, and Fourier transform to obtain a single-channel frequency-domain signal; calculating late reverberation weighting coefficients using the single-channel frequency-domain signal of the previous D frames; estimating the posterior signal-to-noise ratio (SNR) and prior SNR based on the single-channel frequency-domain signal and the corresponding late reverberation weighting coefficients; calculating the probability of speech presence in the current frame using the enhanced speech signal of the previous frame and the frequency-domain signal of the previous frame; calculating gain control parameters using the gain control parameters of the previous frame and the probability of speech presence in the current frame; calculating the gain function based on the prior and posterior SNRs; ​​estimating the enhanced speech audio-domain signal using the gain control parameters, the gain function, and the single-channel frequency-domain signal; comparing the enhanced speech audio-domain signal with a set threshold and updating it according to the comparison result; performing an inverse Fourier transform to convert it to the time domain and outputting it.

Description

Technical Field

[0001] This invention belongs to the field of speech signal processing technology, and particularly relates to a method and apparatus for single-channel de-reverberation of speech signals, a terminal, and a readable storage medium. Background Technology

[0002] In applications involving acoustic equipment, speech reverberation is unavoidable. Reverberation not only reduces the accuracy of speech recognition but also has various negative impacts in video conferencing, voice calls, and marine acoustics. When reverberation is severe, various acoustic devices may even malfunction. Therefore, eliminating reverberation is a crucial issue for related applications. In the paper "A Single-Channel Speech De-reverberation Device" (Publication No.: CN201220287686.3), the proposed solution uses an exponential decay model to estimate the power spectrum of late-stage reverberation and then combines it with spectral subtraction to eliminate late-stage reverberation. The drawback of this solution is that it directly uses an exponential decay model to simulate the Room Impulse Response (RIR) to estimate late-stage reverberation. This simulation method is too crude and simplistic, resulting in inaccurate estimates of late-stage reverberation. Furthermore, directly using spectral subtraction to eliminate reverberation inevitably introduces speech distortion and musical noise. In the paper "A Method and Apparatus for De-reverberation of Speech Data" (Publication No.: CN 201510401640.8), the proposed scheme uses an exponential decay model to estimate the late cross-power spectrum, then combines it with coefficientd spectral subtraction to eliminate late reverberation, and controls the degree of reverberation elimination by adjusting the spectral subtraction coefficient. The drawback of this scheme is that it also directly uses an exponential decay model to simulate the RIR (Room Impulse Response) to estimate late reverberation. This simulation method is too crude and simplistic, resulting in inaccurate estimates of late reverberation. Furthermore, although coefficientd spectral subtraction is added, this parameter can only be manually adjusted, and in practical applications, problems of insufficient or excessive reverberation elimination still exist.

[0003] In summary, there is currently a lack of a method that can accurately estimate late-stage reverberation and effectively eliminate it. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, terminal, and readable storage medium for single-channel dereverberation of speech signals, to solve the technical problem in related technologies that cannot accurately estimate late-stage reverberation and effectively eliminate it. This application estimates the power spectrum of late-stage reverberation by utilizing the correlation of inter-frame speech signals in the frequency domain, and fully utilizes information such as the amplitude and phase of the signal in the spectrum to calculate gain control parameters to eliminate reverberation, thereby suppressing reverberation in single-channel speech signals.

[0005] According to a first aspect of the embodiments of this application, a single-channel de-reverberation method for speech signals is provided, comprising:

[0006] The input single-channel time-domain signal is acquired, and the time-domain signal is subjected to frame segmentation, windowing, and Fourier transform to obtain a single-channel frequency-domain signal.

[0007] Using the single-channel frequency domain signal from the first D frames, calculate the late reverberation weighting coefficient for the i-th frame;

[0008] Based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient, estimate the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame;

[0009] The probability of speech presence in the current frame is calculated using the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame.

[0010] The gain control parameters for the current frame are calculated using the gain control parameters of the previous frame and the speech presence probability of the current frame.

[0011] Based on the amplitude MMSE criterion, the gain function is calculated according to the prior signal-to-noise ratio and the posterior signal-to-noise ratio of the current frame;

[0012] Using the gain control parameters, the gain function, and the single-channel frequency domain signal, the enhanced speech domain signal is estimated;

[0013] The enhanced speech domain signal is compared with a set threshold and updated based on the comparison result;

[0014] Perform an inverse Fourier transform on the updated speech domain signal to convert it to the time domain and output it.

[0015] Furthermore, using the single-channel frequency domain signal from the previous D frames, the late reverberation weighting coefficient of the i-th frame is calculated, including:

[0016] Using the single-channel frequency domain signal from the first D frames, calculate the late reverberation correlation coefficient for each frequency point in the i-th frame:

[0017]

[0018] in, L represents the late reverberation correlation coefficient at the k-th frequency point in the i-th frame, where L is the scaling factor. This represents the single-channel frequency domain signal of the l-th frame, where * indicates taking the conjugate. This indicates modulo operation, where K is the number of frequency points in a frame of frequency domain signal.

[0019] Using the late reverberation correlation coefficients of the past N frames, calculate the late reverberation weighting coefficients for each frequency point in the i-th frame:

[0020]

[0021] in is the late reverberation weighting coefficient for the k-th frequency point in the i-th frame.

[0022] Further, based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficients, the posterior signal-to-noise ratio and the prior signal-to-noise ratio are estimated, including:

[0023] Based on the late reverberation weighting coefficients of the past N frames, estimate the late reverberation power of the i-th frame. ;

[0024] Dividing the single-channel frequency domain power of the i-th frame by the late reverberation power of the i-th frame yields the posterior signal-to-noise ratio of the i-th frame. ;

[0025] The a priori signal-to-noise ratio of the i-th frame is obtained by comparing the posterior signal-to-noise ratio with the lower limit of the posterior signal-to-noise ratio.

[0026] Furthermore, using the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame, the probability of speech presence in the current frame is calculated as follows:

[0027] The probability of speech presence in the current frame is calculated by dividing the sum of the enhanced speech signal power at each frequency point in the (i-1)th frame by the sum of the frequency domain signal power at each frequency point in the same frame.

[0028] Furthermore, using the gain control parameters of the previous frame and the speech presence probability of the current frame, the gain control parameters of the current frame are calculated, where the gain control coefficient of the i-th frame is... The calculation formula is as follows:

[0029]

[0030] in, Represents the smoothing coefficient. and express The maximum and minimum values, Let be the probability of the speech in the i-th frame.

[0031] Furthermore, using the gain control parameters, gain function, and the single-channel frequency domain signal, the enhanced speech domain signal is estimated as follows:

[0032] The enhanced speech domain signal is obtained by multiplying the gain control parameters, gain function, and single-channel frequency domain signal of the i-th frame.

[0033] Further, the enhanced speech domain signal is compared with a set threshold and updated based on the comparison result, including:

[0034] Compare the enhanced audio domain signal of the i-th frame with the set threshold;

[0035] If the enhanced audio domain signal of the i-th frame is greater than the set threshold, then the enhanced audio domain signal of the i-th frame is used as the updated audio domain signal of the i-th frame.

[0036] If the enhanced audio domain signal of the i-th frame is less than or equal to the set threshold, then the single-channel frequency domain signal of the i-th frame is multiplied by a predetermined value to obtain the updated audio domain signal of the i-th frame.

[0037] According to a second aspect of the embodiments of this application, a single-channel de-reverberation device for speech signals is provided, comprising:

[0038] The acquisition module acquires the input single-channel time-domain signal via voice, performs frame segmentation, windowing, and Fourier transform on the time-domain signal to obtain the single-channel frequency-domain signal;

[0039] The first calculation module uses the single-channel frequency domain signal of the previous D frames to calculate the late reverberation weighting coefficient of the i-th frame.

[0040] The first estimation module estimates the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient.

[0041] The second calculation module uses the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame to calculate the probability of speech presence in the current frame.

[0042] The third calculation module uses the gain control parameters of the previous frame and the probability of speech presence in the current frame to calculate the gain control parameters of the current frame.

[0043] The fourth calculation module calculates the gain function based on the amplitude MMSE criterion and the prior and posterior signal-to-noise ratios of the current frame.

[0044] The second estimation module is used to estimate the enhanced speech domain signal using the gain control parameters, the gain function, and the single-channel frequency domain signal.

[0045] The update module is used to compare the enhanced speech domain signal with a set threshold and update it according to the comparison result;

[0046] The conversion module is used to perform an inverse Fourier transform on the updated speech domain signal to convert it to the time domain and output it.

[0047] According to a third aspect of the embodiments of this application, a terminal is provided, comprising:

[0048] One or more processors;

[0049] Memory, used to store one or more programs;

[0050] When the one or more programs are executed by the one or more processors, the one or more processors perform the method as described in the first aspect.

[0051] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0052] The technical solutions provided by the embodiments of this application may include the following beneficial effects:

[0053] As can be seen from the above embodiments, this application uses the correlation of inter-frame speech signals in the frequency domain to estimate the power spectrum of late reverberation, making full use of the amplitude and phase information of the signal in the spectrum; it combines the posterior signal-to-noise ratio based on late reverberation, the prior signal-to-noise ratio and the amplitude MMSE estimator to calculate the gain function; and it uses the speech presence probability to calculate the gain control parameters.

[0054] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0055] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0056] Figure 1 This is a flowchart illustrating a single-channel de-reverberation method for a speech signal according to an exemplary embodiment.

[0057] Figure 2 This is a block diagram illustrating a single-channel de-reverberation device for a speech signal according to an exemplary embodiment.

[0058] Figure 3 This is a schematic diagram of a terminal according to an exemplary embodiment. Detailed Implementation

[0059] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0060] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0061] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0062] Figure 1 This is a flowchart illustrating a single-channel dereverberation method for speech signals according to an exemplary embodiment, such as... Figure 1 As shown, the method may include the following steps:

[0063] Step S11: Obtain the input single-channel time-domain signal, and perform frame segmentation, windowing, and Fourier transform on the time-domain signal to obtain the single-channel frequency-domain signal;

[0064] Step S12: Calculate the late reverberation weighting coefficient of the i-th frame using the single-channel frequency domain signal of the previous D frames;

[0065] Step S13: Estimate the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient;

[0066] Step S14: Calculate the probability of speech presence in the current frame using the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame;

[0067] Step S15: Calculate the gain control parameters for the current frame using the gain control parameters of the previous frame and the speech presence probability of the current frame.

[0068] Step S16: Based on the amplitude MMSE criterion, calculate the gain function according to the prior signal-to-noise ratio and posterior signal-to-noise ratio of the current frame;

[0069] Step S17: Estimate the enhanced speech domain signal using the gain control parameters, gain function, and single-channel frequency domain signal;

[0070] Step S18: Compare the enhanced speech domain signal with a set threshold and update it according to the comparison result;

[0071] Step S19: Perform an inverse Fourier transform on the updated speech domain signal to convert it to the time domain and output it.

[0072] As can be seen from the above embodiments, this application uses the correlation of inter-frame speech signals in the frequency domain to estimate the power spectrum of late reverberation, making full use of the amplitude and phase information of the signal in the spectrum; it combines the posterior signal-to-noise ratio based on late reverberation, the prior signal-to-noise ratio and the amplitude MMSE estimator to calculate the gain function; and it uses the speech presence probability to calculate the gain control parameters.

[0073] In the specific implementation of step S11, the input single-channel time-domain signal is acquired, and the time-domain signal is subjected to frame division, windowing, and Fourier transform to obtain a single-channel frequency-domain signal;

[0074] In one embodiment, single-channel déreverberation is performed on the single-microphone speech signal by first dividing the input time-domain signal into frames to obtain the time-domain signal. , where i represents the i-th frame and n represents the n-th sampling point of that frame. For time-domain signals Performing a Fourier transform converts the signal from the time domain to the frequency domain, as shown in the following expression:

[0075] ,in To represent a window function, you can choose the Hanning window, etc.; k Represents the k-th frequency domain One frequency point.

[0076] In the specific implementation of step S12, the late reverberation weighting coefficient of the i-th frame is calculated using the single-channel frequency domain signal of the previous D frames.

[0077] Specifically, this step may include the following sub-steps:

[0078] Step S21: Using the single-channel frequency domain signal from the previous D frames, calculate the late reverberation correlation coefficient for each frequency point in the i-th frame:

[0079]

[0080] in, L represents the late reverberation correlation coefficient at the k-th frequency point in the i-th frame, where L is the scaling factor. This represents the single-channel frequency domain signal of the l-th frame, where * indicates taking the conjugate. This indicates modulo operation, where K is the number of frequency points in a frame of frequency domain signal. In this example, D=3 and N=20 are taken. The value of L is also assumed to be set and... Here, L=30 is taken; this step utilizes the frequency domain signal of the past D frames and the frequency domain signal of the current frame in step S11 to utilize the correlation between the signals of multiple frames, and uses the late reverberation correlation coefficient calculation formula to obtain the late reverberation correlation coefficient.

[0081] Step S22: Using the late reverberation correlation coefficients of the past N frames, calculate the late reverberation weighting coefficients for each frequency point in the i-th frame:

[0082]

[0083] in is the late reverberation weighting coefficient for the k-th frequency point in the i-th frame.

[0084] In the specific implementation of step S13, the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame are estimated based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient.

[0085] Specifically, this step may include the following sub-steps:

[0086] Step S31: Estimate the late reverberation power at the k-th frequency point in the i-th frame based on the late reverberation weighting coefficients of the past N frames. ;

[0087] Specifically, the expression is ,in This is an adjustment factor that controls the magnitude of the estimated late-stage reverberation power; here, we take... .

[0088] Step S32: Divide the single-channel frequency domain power of the i-th frame by the late reverberation power of the i-th frame to obtain the posterior signal-to-noise ratio of the i-th frame. ;

[0089] Specifically, using the estimated late-stage reverberation power and frequency domain signals Calculate the posterior signal-to-noise ratio .

[0090] Step S33: Compare the posterior signal-to-noise ratio with the lower limit of the posterior signal-to-noise ratio to obtain the prior signal-to-noise ratio of the i-th frame.

[0091] Specifically, using the formula

[0092]

[0093] The prior signal-to-noise ratio was calculated. ,in This is the smoothing coefficient, which can be set to 0.6. It is a posterior signal-to-noise ratio lower limit, which can be set to 0.05.

[0094] In the specific implementation of step S14, the enhanced speech signal of the previous frame and the frequency domain signal of the previous frame are used to calculate the probability of speech presence in the current frame.

[0095] Specifically, the sum of the enhanced speech signal power at each frequency point in the (i-1)th frame is divided by the sum of the frequency domain signal power at each frequency point in the same frame to calculate the speech presence probability in the i-th frame. :

[0096]

[0097] in It is the enhanced speech signal of the (i-1)th frame.

[0098] In the specific implementation of step S15, the gain control parameters of the current frame are calculated using the gain control parameters of the previous frame and the speech presence probability of the current frame.

[0099] Specifically, the gain control coefficient of the i-th frame The calculation formula is as follows:

[0100]

[0101] in, This represents the smoothing coefficient, which can be set to 0.5. and express The maximum and minimum values ​​are set to 1.3 and 0.7 respectively; The initial value is set to 1. Let be the probability of the speech in the i-th frame.

[0102] In the specific implementation of step S16, the gain function is calculated based on the amplitude MMSE criterion and the prior signal-to-noise ratio and posterior signal-to-noise ratio of the current frame.

[0103] Specifically, the gain function G can be calculated using the calculation formula of the amplitude MMSE estimator and the probability distribution model of the speech signal and the late reverberation signal. The probability distribution model can be a complex Gaussian zero-mean distribution or a Laplace distribution, etc. In this embodiment, a complex Gaussian zero-mean distribution is selected, and the speech signal and the late reverberation signal are independent of each other. Therefore, the gain function... The expression is as follows:

[0104]

[0105] in , and Let these represent the zeroth-order Bessel function and the first-order Bessel function, respectively. Here, we consider... ,so and All calculations can be simplified by choosing some approximate expressions.

[0106] In the specific implementation of step S17, the enhanced speech domain signal is estimated using the gain control parameters, the gain function, and the single-channel frequency domain signal.

[0107] Specifically, the gain control parameter, gain function, and single-channel frequency domain signal of the i-th frame are multiplied to obtain the enhanced speech domain signal, i.e.:

[0108]

[0109] In the specific implementation of step S18, the enhanced speech domain signal is compared with a set threshold and updated according to the comparison result;

[0110] Specifically, to prevent the enhanced speech signal from being too weak, a lower limit protection measure is implemented, expressed as follows:

[0111] ,Right now:

[0112] The enhanced speech domain signal of the i-th frame is compared with a set threshold; if the enhanced speech domain signal of the i-th frame is greater than the set threshold, the enhanced speech domain signal of the i-th frame is used as the updated speech domain signal of the i-th frame; if the enhanced speech domain signal of the i-th frame is less than or equal to the set threshold, the single-channel frequency domain signal of the i-th frame is multiplied by a predetermined value to obtain the updated speech domain signal of the i-th frame.

[0113] Among them, the predetermined value It is a small value, and can be set to 0.02.

[0114] In the specific implementation of step S19, the updated speech domain signal is subjected to inverse Fourier transform to convert it to the time domain and then output.

[0115] Specifically, the inverse Fourier transform is used to transform S18 Convert back to the time domain and output.

[0116] Corresponding to the aforementioned embodiments of the single-channel dereverberation method for speech signals, this application also provides embodiments of a single-channel dereverberation device for speech signals.

[0117] Figure 2 This is a block diagram of a single-channel de-reverberation device for speech signals according to an exemplary embodiment. (Refer to...) Figure 2 The device may include:

[0118] The acquisition module 21 acquires the input single-channel time-domain signal via voice, performs frame segmentation, windowing, and Fourier transform on the time-domain signal to obtain a single-channel frequency-domain signal;

[0119] The first calculation module 22 uses the single-channel frequency domain signal of the previous D frames to calculate the late reverberation weighting coefficient of the i-th frame.

[0120] The first estimation module 23 estimates the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient.

[0121] The second calculation module 24 uses the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame to calculate the probability of speech presence in the current frame.

[0122] The third calculation module 25 calculates the gain control parameters of the current frame using the gain control parameters of the previous frame and the probability of speech presence in the current frame.

[0123] The fourth calculation module 26 calculates the gain function based on the amplitude MMSE criterion and the prior signal-to-noise ratio and posterior signal-to-noise ratio of the current frame.

[0124] The second estimation module 27 is used to estimate the enhanced speech domain signal using the gain control parameters, the gain function and the single-channel frequency domain signal;

[0125] Update module 28 is used to compare the enhanced speech domain signal with a set threshold and update it according to the comparison result;

[0126] The conversion module 29 is used to perform an inverse Fourier transform on the updated speech domain signal to convert it to the time domain and output it.

[0127] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0128] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0129] Accordingly, this application also provides a terminal, comprising: one or more processors; a memory for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the single-channel dereverberation method for voice signals as described above. Figure 3 The diagram shown is a hardware structure diagram of any device with data processing capabilities, used in an embodiment of the present invention to provide a single-channel de-reverberation method for speech signals. (Except for...) Figure 3 In addition to the processor, memory, DMA controller, disk, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0130] Accordingly, this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the single-channel dereverberation method for speech signals as described above. The computer-readable storage medium can be an internal storage unit of any data-processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device for a wind turbine, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., mounted on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data-processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data-processing device, and can also be used to temporarily store data that has been output or will be output.

[0131] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0132] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A single-channel de-reverberation method for speech signals, characterized in that, include: The input single-channel time-domain signal is acquired, and the time-domain signal is subjected to frame segmentation, windowing, and Fourier transform to obtain a single-channel frequency-domain signal. Using the single-channel frequency domain signal from the first D frames, calculate the late reverberation weighting coefficient for the i-th frame; Based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient, estimate the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame; The probability of speech presence in the current frame is calculated using the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame. The gain control parameters for the current frame are calculated using the gain control parameters of the previous frame and the speech presence probability of the current frame. Based on the amplitude MMSE criterion, the gain function is calculated according to the prior signal-to-noise ratio and the posterior signal-to-noise ratio of the current frame; Using the gain control parameters, the gain function, and the single-channel frequency domain signal, the enhanced speech domain signal is estimated; The enhanced speech domain signal is compared with a set threshold and updated based on the comparison result; Perform an inverse Fourier transform on the updated speech domain signal to convert it to the time domain and output it.

2. The method according to claim 1, characterized in that, Using the single-channel frequency domain signal from the previous D frames, the late reverberation weighting coefficients for the i-th frame are calculated, including: Using the single-channel frequency domain signal from the first D frames, calculate the late reverberation correlation coefficient for each frequency point in the i-th frame: ， in, L represents the late reverberation correlation coefficient at the k-th frequency point in the i-th frame, where L is the scaling factor. This represents the single-channel frequency domain signal of the l-th frame, where * indicates taking the conjugate. This indicates modulo operation, where K is the number of frequency points in a frame of frequency domain signal. Using the late reverberation correlation coefficients of the past N frames, calculate the late reverberation weighting coefficients for each frequency point in the i-th frame: ， in Let be the late reverberation weighting coefficient for the k-th frequency point in the i-th frame, where i is greater than N.

3. The method according to claim 1, characterized in that, Based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficients, the posterior signal-to-noise ratio and the prior signal-to-noise ratio are estimated, including: Based on the late reverberation weighting coefficients of the past N frames, estimate the late reverberation power of the i-th frame. ; Dividing the single-channel frequency domain power of the i-th frame by the late reverberation power of the i-th frame yields the posterior signal-to-noise ratio of the i-th frame. ; The a priori signal-to-noise ratio of the i-th frame is obtained by comparing the posterior signal-to-noise ratio with the lower limit of the posterior signal-to-noise ratio.

4. The method according to claim 1, characterized in that, Using the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame, the probability of speech presence in the current frame is calculated as follows: The probability of speech presence in the current frame is calculated by dividing the sum of the enhanced speech signal power at each frequency point in the (i-1)th frame by the sum of the frequency domain signal power at each frequency point in the same frame.

5. The method according to claim 1, characterized in that, Using the gain control parameters of the previous frame and the speech presence probability of the current frame, calculate the gain control parameters of the current frame, where the gain control coefficient of the i-th frame is... The calculation formula is as follows: ， in, Represents the smoothing coefficient. and express The maximum and minimum values, Let be the probability of the speech existing in the i-th frame.

6. The method according to claim 1, characterized in that, Using the gain control parameters, gain function, and single-channel frequency domain signal, the enhanced speech domain signal is estimated as follows: The enhanced speech domain signal is obtained by multiplying the gain control parameters, gain function, and single-channel frequency domain signal of the i-th frame.

7. The method according to claim 1, characterized in that, The enhanced speech domain signal is compared with a set threshold and updated based on the comparison result, including: Compare the enhanced audio domain signal of the i-th frame with the set threshold; If the enhanced audio domain signal of the i-th frame is greater than the set threshold, then the enhanced audio domain signal of the i-th frame is used as the updated audio domain signal of the i-th frame. If the enhanced audio domain signal of the i-th frame is less than or equal to the set threshold, then the single-channel frequency domain signal of the i-th frame is multiplied by a predetermined value to obtain the updated audio domain signal of the i-th frame.

8. A single-channel de-reverberation device for speech signals, characterized in that, include: The acquisition module acquires the input single-channel time-domain signal via voice, performs frame segmentation, windowing, and Fourier transform on the time-domain signal to obtain the single-channel frequency-domain signal; The first calculation module uses the single-channel frequency domain signal of the previous D frames to calculate the late reverberation weighting coefficient of the i-th frame. The first estimation module estimates the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the current frame based on the single-channel frequency domain signal and the corresponding late reverberation weighting coefficient. The second calculation module uses the enhanced speech signal from the previous frame and the frequency domain signal from the previous frame to calculate the probability of speech presence in the current frame. The third calculation module uses the gain control parameters of the previous frame and the probability of speech presence in the current frame to calculate the gain control parameters of the current frame. The fourth calculation module calculates the gain function based on the amplitude MMSE criterion and the prior and posterior signal-to-noise ratios of the current frame. The second estimation module is used to estimate the enhanced speech domain signal using the gain control parameters, the gain function, and the single-channel frequency domain signal. The update module is used to compare the enhanced speech domain signal with a set threshold and update it according to the comparison result; The conversion module is used to perform an inverse Fourier transform on the updated speech domain signal to convert it to the time domain and output it.

9. An electronic terminal, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-7.

Images