Reverberation removal method, microphone array chip, microphone array and computer program product

Abstract

A reverberation removal method, a microphone array chip, a microphone array and a computer program product. The reverberation removal method comprises: acquiring information of a filter for performing reverberation removal processing on a first sound signal, and filter parameters of the filter; on the basis of the filter parameters, iteratively updating the filter at least twice to obtain an updated filter; and on the basis of information of the updated filter, determining a filter for performing reverberation removal processing on a second sound signal, wherein the second sound signal is a signal that is adjacent to the first sound signal and temporally later than the first sound signal. The method effectively reduces residual reverberation in a sound signal.

Description

Reverberation removal method, microphone array chip, microphone array and computer program product

[0001] This application claims priority to the Chinese invention patent application with application number 202410330279.3, entitled “Reverberation removal method, microphone array chip, microphone array and computer program product”, filed on March 21, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0002] The embodiments of the present application relate to the field of signal processing technology, and in particular to a reverberation removal method, a microphone array chip, a microphone array, and a computer program product. Background Art

[0003] A microphone array is an array of acoustic sensors (such as microphones) that can be used to sample and process sound signals. The physical environment in which the sound signals are located often contains interfering signals such as reverberation and echo. Reverberation is the sound emitted by a sound source that reflects off physical surroundings, such as walls or objects in a room, and then travels back to the microphone. This reverberation is then superimposed on the direct sound from the target sound source and received by the acoustic sensor. Consequently, the microphone array processes the sound signal based on this superimposed sound signal, interfering with subsequent processing such as sound source localization and speech enhancement.

[0004] Unlike echo, although both are interference signals, echo is a single, fixed reflection whose intensity and time delay are sufficient to distinguish it from the target sound source, making it easier to remove. Reverberation, on the other hand, is formed by multiple back-and-forth diffuse reflections, resulting in a mixture of reflections arriving at different angles and times, which gradually decay and are difficult to remove.

[0005] To this end, a related art approach, called WPE (Weighted Prediction Error), provides an approach. This approach first estimates the reverberation tail of the sound signal and then subtracts it from the observed signal. This yields a maximum likelihood estimate of the weak reverberation signal, which is then used as the dereverberation result. However, this approach is not very effective in dereverberation, and residual reverberation may still exist in the sound signal.

[0006] Therefore, how to remove the reverberation signal from the sound signal and avoid the influence of residual reverberation as much as possible has become an urgent problem to be solved. Summary of the Invention

[0007] In view of this, an embodiment of the present application provides a reverberation removal solution to at least partially solve the above-mentioned problem.

[0008] According to a first aspect of an embodiment of the present application, a dereverberation method is provided, comprising: obtaining information about a filter for performing dereverberation processing on a first sound signal and filter parameters of the filter; iteratively updating the filter at least twice based on the filter parameters to obtain an updated filter; and determining, based on the information about the updated filter, a filter for performing dereverberation processing on a second sound signal, wherein the second sound signal is a signal that is adjacent to the first sound signal and is later in time than the first sound signal.

[0009] According to a second aspect of an embodiment of the present application, a chip for a microphone array is provided, wherein the chip includes a processor, and the processor is configured to execute the method described in the first aspect.

[0010] According to a third aspect of an embodiment of the present application, a microphone array is provided, comprising the microphone array chip as described in the second aspect.

[0011] According to a fourth aspect of the embodiments of the present application, a computer storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method described in the first aspect is implemented.

[0012] According to a fifth aspect of an embodiment of the present application, a computer program product is provided, comprising computer instructions, wherein the computer instructions instruct a processor to perform operations corresponding to the method described in the first aspect.

[0013] According to the solution provided by the embodiment of the present application, when the sound signal is subjected to dereverberation processing, the filter used for the preceding sound signal, i.e., the first sound signal, is updated and then selected for use by the adjacent subsequent sound signal, i.e., the second sound signal. Unlike the traditional filter which is updated once and then handed over for subsequent use, in the solution of the embodiment of the present application, the filter parameters of the filter are updated at least twice based on the filter, and the final updated filter is determined by the iterative update result and then handed over for subsequent use. That is, for the adjacent first and second sound signals, the traditional method uses the filter to perform dereverberation processing on the first sound signal, and then it is used as a candidate filter for the second sound signal after one filter update; while in the solution of the present application, after the filter is used to perform dereverberation processing on the first sound signal, it is updated N times (N is greater than or equal to 2) and then used as a candidate filter for the second sound signal. Thus, the filter parameters are optimized as much as possible, so that the filter performance is no longer affected by the filter update speed or update frequency, which not only ensures the performance of the filter, but also ensures the robustness of the filter, thereby effectively reducing the residual reverberation in the sound signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0015] FIG1 is a schematic diagram of the structure of an exemplary microphone array applicable to an embodiment of the present application;

[0016] FIG2A is a flowchart of steps of a reverberation removal method according to an embodiment of the present application;

[0017] FIG2B is a schematic diagram of a pseudo-code-based dereverberation process in the embodiment shown in FIG2A ;

[0018] FIG2C is a schematic diagram of an example scenario in the embodiment shown in FIG2A ;

[0019] FIG3 is a structural block diagram of a chip for a microphone array according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field should fall within the scope of protection of the embodiments of the present application.

[0021] The specific implementation of the embodiment of the present application is further explained below in conjunction with the accompanying drawings of the embodiment of the present application.

[0022] Figure 1 illustrates an exemplary microphone array applicable to the embodiments of the present application. As shown in Figure 1 , the microphone array may include multiple acoustic sensors and a chip 104. In this example, the multiple acoustic sensors are exemplified by multiple microphones 102, which can adopt any topological structure and have any directional properties. In other words, the embodiments of the present application do not limit the specific implementation of the microphone array, and the embodiments of the present application are applicable to all devices.

[0023] The plurality of microphones 102 are used to collect sound signals in the physical space where the microphone array is located; the chip 104 is used to process the sound signals collected by the plurality of microphones 102 , including but not limited to reverberation removal (ie, dereverberation) processing.

[0024] In an exemplary dereverberation method, the chip 104 may perform dereverberation processing on the sound signal using a WPE method. Furthermore, the chip 104 in this example may perform dereverberation processing on the sound signal based on a SwWPE (Switching WPE) method.

[0025] Unlike traditional WPE, the SwWPE method uses multiple (two or more) linear prediction filters. When performing dereverberation processing on a sound signal in a certain frequency band at a certain moment, a better filter is selected from these multiple linear prediction filters to perform dereverberation processing on the sound signal. The selected filter is obtained by performing a single filter update on the filter after performing dereverberation processing on the sound signal at the previous moment.

[0026] While the SwWPE method improves dereverberation to a certain extent compared to the WPE method, it is based on RLS (Recursive Least Squares) implementation. RLS's performance is limited by the forgetting factor λ in its algorithm. A larger value for the forgetting factor λ results in a faster filter update speed. However, the robustness of the filter varies inversely with the update speed. That is, faster filter updates result in poor robustness. Consequently, dereverberation filtering is less effective.

[0027] To this end, the embodiment of the present application provides a dereverberation solution to avoid the above phenomenon and improve the dereverberation effect of the filter. The dereverberation solution of the embodiment of the present application is specifically described below through an embodiment.

[0028] 2A , there is shown a flowchart of steps of a reverberation removal method according to an embodiment of the present application.

[0029] The reverberation removal method of this embodiment includes the following steps:

[0030] Step S202: Acquire information about a filter for performing dereverberation processing on the first sound signal and filter parameters of the filter.

[0031] For a microphone array, dereverberation processing needs to be performed on a sound signal, such as a sound signal (i.e., a first sound signal) collected by a microphone in a certain frequency band at time t (also called time frame t) and containing an interference signal. This first sound signal can also be called an observed signal. As mentioned above, dereverberation processing of sound signals can be achieved through filters.

[0032] For the convenience of description, in the embodiments of this application, Characterizing the selected filter, in the SwWPE mode, the target sound signal obtained after dereverberation processing by the filter can be expressed as follows:

[0033] in, represents the target sound signal estimation after dereverberation; represents the observation signal collected by the mth microphone (the total number of microphones is M) on the frequency band ω at time t, which contains the interference signal; c * Indicates the sequence number of the filter selected from multiple filters; represents the selected filter; H represents the conjugate transpose; y t-D，ω Represents the sound signal vector collected by multiple microphones under predetermined time delay.

[0034] in, In the formula, D represents a predetermined time delay, D>0; M represents the number of microphones in the microphone array; L represents the filter length; and T represents transposition.

[0035] In one feasible manner, the multiple filters may be adaptive filters based on the RLS (Recursive Least Squares) algorithm. This filter is based on the least squares method and recursively updates the filter coefficients to minimize the mean square error of the prediction error, thereby achieving real-time estimation and filtering of the signal. Based on this, in one feasible manner, the filters selected above are It can be obtained based on the calculation results of the mean square error of the prediction errors corresponding to multiple filters. For example, the filter with the smallest mean square error is selected as the filter. Gets information about the filter that performs dereverberation processing.

[0036] In the SwWPE method, the selected filter pair After the dereverberation process is performed, multiple filters including the filter need to be updated, and the next time a filter is selected, the selection is made based on the updated multiple filters. In other words, the currently selected filter is also the filter that was updated after the previous dereverberation process.

[0037] Still with filter For example, the acquisition process can be expressed as:

[0038] In formula 2, represents the filter at time t-1 before time t; k t,ω represents the Kalman gain; e0 represents the prior error in the RLS algorithm; H represents the conjugate transpose.

[0039] Then, in a specific example, this step obtains information about a filter for performing dereverberation processing on the first sound signal and filter parameters of the filter, which can be exemplarily implemented as obtaining Information, and Corresponding to the above k t,ω and e0, as well as the parameters involved in the calculation process of these parameters, etc.

[0040] Step S204: Based on the filter parameters of the filter, perform at least two iterative updates on the filter to obtain an updated filter.

[0041] As mentioned above, in the SwWPE method, the filter needs to be updated before being used for the sound signal at the next moment. For example, as shown in the above formula 2.

[0042] Furthermore, in one example, in the filter parameters shown in Formula 2:

[0043] The meanings of the parameters in Formula 3 can be referred to the above explanations and will not be repeated here.

[0044] in, represents the state covariance matrix of the prior estimate, which is used to provide a measure of the uncertainty of the current state estimate of the filter; κ represents the adaptive weight, Among them, p∈[0,2], |||||2 represents the L2 norm; Represents the target sound signal after dereverberation; other parameters can refer to the above explanations.

[0045] Furthermore,

[0046] Where λ (0 < λ < 1) represents the forgetting factor. The larger λ is, the slower the update speed of Φ is and the better the robustness of the filter is. Conversely, the smaller λ is, the faster the update speed of Φ is and the worse the robustness of the filter is.

[0047] By using the Kalman gain, when the observed signal is noisy, the Kalman gain is reduced, reducing the weight of the observed signal; and when the observed signal is noisy, the Kalman gain is increased, increasing the weight of the observed signal. Therefore, by updating the Kalman gain, the filter can be more resistant to noise and achieve a better dereverberation effect.

[0048] To facilitate explanation of the filter update, in the embodiment of the present application, the following formula 6 is used to replace the aforementioned formula 2, and the filter update is expressed as:

[0049] in,

[0050] Then, after the sound signal at time t-1 is de-reverberated, e0 is updated as follows:

[0051] The above e1 will be used as the posterior error output of the RLS adaptive filter to be used for sound signal processing at the next moment.

[0052] As can be seen, in this approach, the filter is only updated once for two adjacent sound signals. Moreover, due to the setting of λ in the Kalman gain update, the filter update cannot take into account both update speed and robustness.

[0053] Based on this, unlike the filter that is only updated once in the aforementioned formulas 6 and 7, in the embodiment of the present application, the filter that processes the first sound signal is iteratively updated multiple times (two or more times) after it performs dereverberation processing on the first sound signal, in order to optimize the filter as much as possible based on the existing filter parameters, so that it can be updated quickly while also having good robustness. In the embodiment of the present application, the multiple iterative updates of the filter are based on the iterative updates of the aforementioned filter parameter e0.

[0054] Specifically, in the RLS adaptive filter, the filter is iteratively updated at least twice based on the filter parameters to obtain the updated filter. This can be implemented as follows: Based on the prior error, the RLS adaptive filter is iteratively updated at least twice to obtain the updated filter. The RLS adaptive filter can automatically adjust the filter parameters based on the characteristics of the input signal using the RLS algorithm, effectively suppressing noise and extracting signal features.

[0055] The process can be schematically illustrated as follows:

[0056] Where N represents the number of iterations; h N represents the filter after N iterations; h N-1 represents the filter after the N-1th iteration; k t,ω represents the Kalman gain; e N-1 Represents the posterior error output of the N-1th iteration.

[0057] The difference from the above formula 7 is that the above e N It will be output as the a posteriori error after N iterations of the RLS adaptive filter to be used for processing the sound signal (second sound signal) at the next moment.

[0058] As can be seen, in this approach, the filter is iteratively updated N times for two adjacent sound signals. Furthermore, because the Kalman gain is not involved in these N iterations, the filter is no longer restricted by the setting of λ. This allows for a higher update speed and greater robustness through multiple iterations.

[0059] Although the above process can realize the iterative update of the filter, in order to simplify the calculation and maintain consistency with the SwWPE calculation method, in a feasible way, the update of the above filter can be expressed as:

[0060] Among them, β represents the update coefficient of the filter, represents the updated coefficient of the filter on the frequency band ω at time t, and α t，ω ∈[0, 1), β is greater than or equal to 1.

[0061] It can be seen that when obtaining the update coefficient β, the Kalman gains corresponding to multiple observation signals can be obtained, where the multiple observation signals are collected by multiple microphones in the microphone array; based on the Kalman gain and the Nth power result of the Kalman gain, the update coefficient β is obtained, where N is equal to the number of iterative updates of the RLS adaptive filter.

[0062] Based on this, the update of the RLS adaptive filter can be implemented as follows: based on the prior error, the RLS adaptive filter is iteratively updated at least twice according to the update coefficient β used to indicate the update frequency to obtain an updated filter; wherein the update coefficient is greater than or equal to 1.

[0063] Therefore, during the filter update process, the t,ω The iterative update of the filter can be realized, and the additional computational effort required is very small and almost negligible, effectively realizing the iterative update of the filter under low computational effort conditions.

[0064] Step S206: Based on the updated filter information, determine a filter for performing dereverberation processing on the second sound signal.

[0065] The second sound signal is a signal that is adjacent to the first sound signal and is later in time than the first sound signal.

[0066] In one feasible approach, the updated filter can be used to dereverberate the second sound signal. That is, the updated filter is determined as the filter for dereverberation of the second sound signal, and the updated filter is used to dereverberate the second sound signal to obtain the target sound signal. As previously mentioned, a filter that is iteratively updated multiple times not only has good robustness but also can have a faster update speed, thus also having good convergence and better overall performance. Therefore, this filter can be used to process the next sound signal adjacent to the first sound signal, namely the second sound signal, improving the efficiency of filter selection and determination.

[0067] However, based on the multiple filter selection strategies under the SwWPE method, and avoiding the spatial non-stationarity of background noise capture that may be generated by the fixed use of a certain filter, when determining the filter for the second sound signal, it is still possible to select from multiple filters including the updated filter. In this case, in addition to the iteratively updated filter, the scheme of the embodiment of the present application will also perform parameter updates on the other filters in the multiple filters except the updated filter. This parameter update mainly includes the update of the filter parameters used to calculate the Kalman gain, for example, the update of Φ. Due to the need to calculate the Kalman gain, the updated filter will also update its own Φ, for example, it can be updated with reference to Formula 5.

[0068] The parameter updates for other filters can be implemented as follows:

[0069] Where c represents the filter number, c = 1, ..., C, C represents the number of filters, and c ≠ c * , θ is a constant between 0 and 1.

[0070] The above process is described below in pseudo code form with reference to Figure 2B. In Figure 2B, the numbers on the left represent pseudo code line numbers.

[0071] As can be seen from FIG2B , the input of the dereverberation process in the embodiment of the present application is the observation signal (sound signal at time t) and y t-D，ω , the output is the target sound signal after dereverberation The first line illustrates the initialization process of multiple filters and the prior estimated state covariance matrix Φ corresponding to each filter. On this basis, lines 2-14 illustrate the dereverberation processing and filter update process of the sound signal at different time t through a for loop. Among them, the third line illustrates the sequence number c of the filter selected from C filters for dereverberation processing of the sound signal at the current time, i.e., the first sound signal. *The 4th row shows the process of obtaining the prior error e0. The 5th to 7th rows show the process of obtaining the parameters for calculating the Kalman gain and obtaining the Kalman gain based on the obtained parameters. The 8th and 9th rows show the process of obtaining the filter update coefficient β. The 10th row shows the process of obtaining the filter number c * The filter is iteratively updated to obtain the updated filter Line 11 shows the use of the updated filter For sound signals Processing to obtain the target sound signal The 12th line shows the updated filter The process of updating the filter parameters (the state covariance matrix Φ estimated a priori). Line 13 shows the process of updating the filter parameters after the update. The process of updating the filter parameters (the prior estimated state covariance matrix Φ) of other filters except .

[0072] Through the above process, both the convergence and robustness of the filter are guaranteed, and the reverberation removal effect of the filter is guaranteed.

[0073] To further illustrate the dereverberation effect of the embodiment of the present application, a comparison is made below with a conventional SwWPE method.

[0074] First, substitute h1 in Formula 6 into Formula 7 to obtain:

[0075] From this, we can deduce the e in formula nine N It can be expressed as:

[0076] In the conventional SwWPE method, the output is the a posteriori error, i.e., e1. Based on this, Formula 13 can be viewed as: SwWPE multiplied by a post-gain w post , the post-gain w post Right now That is,

[0077] Combined with the Kalman gain of formula 4, w post It can be expressed as:

[0078] in, is a positive definite Hermitian matrix, is a real number, so the above formula 14 can be expressed as:

[0079] in, Considering can be considered as an estimate In the case of the RLS covariance matrix, ε can be interpreted as the signal-to-reverberation ratio. Therefore, w post ∈[0,1) is considered as a variant of the Wiener filter with a power of N-1.

[0080] It can be seen that the dereverberation solution of the embodiment of the present application applies a post-gain similar to that of the Wiener filter to the residual reverberation suppression of the conventional SwWPE, and therefore has a better dereverberation effect than the conventional SwWPE.

[0081] The above process is exemplified below using a scenario example, as shown in FIG2C .

[0082] In Figure 2C, assume the microphone array shown in Figure 1 is used, with four microphones. User A is in a conference room and speaks using the microphone array. The microphone array uses four microphones to capture user A's voice, but because it is in a conference room, it also captures interference signals including reverberation and echo. To address the reverberation, the microphone array uses a chip within it to perform dereverberation processing. Furthermore, assume that the chip uses four filters for dereverberation processing.

[0083] Taking the dereverberation processing of the sound signal of a certain frequency band collected by the first microphone at time t (simply represented as sound signal 1) as an example, the chip will select a filter with better performance from the four filters and set it as the second filter, recorded as filter 2. If filter 2 is selected for the first time, the sound signal 1 is dereverberated based on its initialization filter parameters to obtain the target sound signal 1' after dereverberation. Furthermore, the chip will determine the corresponding update coefficient β for filter 2 in the manner shown in formula 10. Filter 2 is then iteratively updated based on the update coefficient β. In this example, it is set to 8 iterations. After 8 iterative updates, the updated filter 2 is obtained. In addition, the parameters Φ used to calculate the Kalman gain in filter 2 and the other 3 filters will also be updated to achieve the update of all 4 filters.

[0084] At time t+1, the first microphone collects the sound signal 2 of this frequency band, and will again compare which of the four updated filters has better performance, so as to select the filter for dereverberation processing of the sound signal 2. Assuming that filter 2 is selected again, filter 2 will use the filter parameters updated after 8 iterations to perform reverberation processing on the sound signal 2, and obtain the target sound signal 2' after dereverberation. Then, the chip will re-determine the corresponding update coefficient β for the filter 2 in the manner shown in Formula 10. Based on the re-determined update coefficient β, the filter 2 is iteratively updated again 8 times. Furthermore, the parameter Φ used to calculate the Kalman gain in the filter 2 and the other 3 filters will also be updated again.

[0085] If, at time t+2, the first microphone captures the sound signal 3 in this frequency band, it will again compare the four updated filters for better performance, selecting the filter to dereverberate the sound signal 3. Assuming that filter 3 is selected this time, reference will be made to filter 2, which processed sound signal 1 above. If this filter 3 is selected for the first time, dereverberation will be performed on the sound signal 3 based on its initialized filter parameters, obtaining the dereverberated target sound signal 3'. If filter 3 has been selected before and iteratively updated, this filter 3 will dereverberate the sound signal 3 based on the filter parameters from the previous iterative update, obtaining the dereverberated target sound signal 3'.

[0086] The above process is repeated in a loop, and each selected filter is iteratively updated for subsequent use until all sound signals are processed.

[0087] As can be seen from the above, through the solution provided by the embodiment of the present application, when the sound signal is subjected to dereverberation processing, the filter used for the preceding sound signal, i.e., the first sound signal, is updated and then selected for use by the adjacent subsequent sound signal, i.e., the second sound signal. Unlike the traditional filter that is updated once and then handed over for subsequent use, in the embodiment of the present application, the filter parameters of the filter are updated at least twice based on the filter, and the final updated filter is determined by the iterative update result and then handed over for subsequent use. That is, for the adjacent first and second sound signals, the traditional method uses the filter to perform dereverberation processing on the first sound signal, and then it is used as a candidate filter for the second sound signal after one filter update; while in the solution of the present application, after the filter is used to perform dereverberation processing on the first sound signal, it is updated N times (N is greater than or equal to 2) and then used as a candidate filter for the second sound signal. Thus, the filter parameters are optimized as much as possible, so that the filter performance is no longer affected by the filter update speed or frequency, which not only ensures the performance of the filter, but also ensures the robustness of the filter, thereby effectively reducing the residual reverberation in the sound signal.

[0088] In an embodiment of the present application, a chip for a microphone array is also provided, as shown in Figure 3. In Figure 3, the chip includes at least a processor, which can obtain sound signals collected by acoustic sensors such as microphones in the microphone array and use the reverberation removal method described above to dereverberate the sound signals.

[0089] On this basis, the chip in the microphone array shown in FIG1 may adopt the chip form shown in FIG3 above to perform efficient dereverberation processing of the sound signal.

[0090] In addition, an embodiment of the present application further provides a computer program product, including computer instructions, which can instruct a processor to perform operations corresponding to the above-mentioned reverberation removal method.

[0091] An embodiment of the present application further provides a computer storage medium having a computer program stored thereon, which implements the aforementioned reverberation removal method when executed by a processor.

[0092] The computer program product / computer storage medium embodiment in the embodiment of the present application has been described in detail in the aforementioned reverberation removal method embodiment. Therefore, its relevant content and beneficial effects can be understood with reference to the aforementioned method embodiment and will not be repeated here.

[0093] In addition, it should be noted that the user-related information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to sample data used to train the model, data used for analysis, stored data, displayed data, user voice data, etc.) involved in the embodiments of this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0094] It should be pointed out that, according to the needs of implementation, the various components / steps described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0095] The above-mentioned method according to the embodiment of the present application can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded via a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor or programmable or dedicated hardware (such as an application-specific integrated circuit (ASIC) or a field programmable gate array (FPGA)). It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component (e.g., random access memory (RAM), read-only memory (ROM), flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown here, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown here.

[0096] Those skilled in the art will appreciate that the units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for specific applications, but such implementation should not be considered to be beyond the scope of the embodiments of this application.

[0097] The above implementation methods are only used to illustrate the embodiments of the present application, and are not intended to limit the embodiments of the present application. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the embodiments of the present application. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of the present application, and the scope of patent protection of the embodiments of the present application should be defined by the claims.

Claims

1. A reverberation removal method, comprising: Acquiring information about a filter that performs dereverberation processing on the first sound signal and filter parameters of the filter; Based on the filter parameters, performing at least two iterative updates on the filter to obtain an updated filter; Based on the information of the updated filter, a filter for performing dereverberation processing on a second sound signal is determined, wherein the second sound signal is a signal adjacent to the first sound signal and later in time than the first sound signal.

2. The method according to claim 1, wherein The filter is a recursive least squares RLS adaptive filter, and the filter parameters include a priori errors of the RLS adaptive filter; The step of iteratively updating the filter at least twice based on the filter parameters to obtain an updated filter comprises: Based on the priori error, the RLS adaptive filter is iteratively updated at least twice to obtain an updated filter.

3. The method according to claim 2, wherein: The step of iteratively updating the RLS adaptive filter at least twice based on the prior error to obtain an updated filter comprises: Based on the priori error, the RLS adaptive filter is iteratively updated at least twice according to an update coefficient indicating an update frequency to obtain an updated filter; wherein the update coefficient is greater than or equal to 1.

4. The method according to claim 3, wherein: The update coefficient is obtained by: Obtaining Kalman gains corresponding to a plurality of observation signals, wherein the plurality of observation signals are collected by a plurality of microphones in a microphone array; The update coefficient is obtained based on the Kalman gain and the N-th power result of the Kalman gain, wherein N is equal to the number of iterative updates of the RLS adaptive filter.

5. The method according to any one of claims 1 to 4, wherein: The obtaining information of a filter for performing dereverberation processing on the first sound signal includes: From a plurality of preset filters, a filter with the smallest mean square error is determined as a filter for performing dereverberation processing on the first sound signal, and information of the determined filter is acquired.

6. The method according to claim 5, wherein: The step of determining a filter for performing dereverberation processing on the second sound signal based on the information of the updated filter comprises: The updated filter is determined as a filter for performing dereverberation processing on the second sound signal, and the updated filter is used to perform dereverberation processing on the second sound signal to obtain a target sound signal.

7. The method according to claim 6, wherein: After obtaining the target sound signal, the method further includes: Parameters of filters used for calculating Kalman gains in the multiple filters are updated.

8. A computer program product, comprising computer instructions, wherein the computer instructions instruct a processor to execute operations corresponding to the method according to any one of claims 1 to 7.

9. A chip for a microphone array, comprising a processor, wherein the processor is configured to execute the method according to any one of claims 1 to 7.

10. A microphone array, comprising the microphone array chip as claimed in claim 9.

11. A computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images