Filter model updating method, device, equipment, medium and computer program product

Abstract

The present application discloses a filter model update method, device, equipment, medium and computer program product. The method includes obtaining a wake-up score of a first speech frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first speech frame signal contains the wake-up word; calculating a wake-up trigger value according to a wake-up trigger threshold and a wake-up score mean, wherein the wake-up score mean is a mean value calculated according to the wake-up score distribution of historical speech frame signals of the wake-up word in a historical wake-up speech set; when the wake-up score is less than the wake-up trigger threshold, determining a reduction factor of the update step of the filter model according to the relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value; and updating the filter model according to the update step and the reduction factor. According to the embodiments of the present application, the accuracy of the filter model update can be improved.

Description

Technical Field

[0001] The present application belongs to the field of speech enhancement technology, and in particular, relates to a filter model updating method, device, equipment, medium and computer program product. Background Art

[0002] With the advent of the mobile Internet and artificial intelligence era, voice interaction has achieved unprecedented growth in recent years. Among them, voice wake-up technology, as a special voice recognition technology, has become an important part of the interaction between users and machines.

[0003] At present, voice wake-up technology has undergone several rounds of optimization and iteration. In actual application scenarios, before judging whether the current voice signal can wake up the device, the current voice frame signal will be filtered through the filter model to remove invalid and interfering voices such as noise. The filter model will automatically update and iterate while filtering the voice frame signal, but it is still facing the problem of poor wake-up effect caused by unsatisfactory filtering results of the filter model.

[0004] Therefore, how to improve the accuracy of filtering model updating has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The embodiments of the present application provide a filter model updating method, apparatus, device, medium and computer program product to solve the problem of how to reduce the damage to speech caused by filter coefficient updating, and can improve the accuracy of filter model updating.

[0006] In a first aspect, an embodiment of the present application provides a filtering model updating method, comprising:

[0007] Obtaining a wake-up score of a first voice frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first voice frame signal contains the wake-up word;

[0008] Calculate the wake-up trigger value according to the wake-up trigger threshold and the wake-up score mean, wherein the wake-up score mean is the mean value calculated according to the wake-up score distribution of the historical speech frame signal of the wake-up word in the historical wake-up speech set;

[0009] When the wake-up score is less than the wake-up trigger threshold, determining a reduction factor of the update step size of the filtering model according to a size relationship among the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value;

[0010] The filtering model is updated according to the update step size and the reduction factor.

[0011] In some embodiments, when the wake-up score is less than the wake-up trigger threshold, determining the reduction factor of the update step size of the filtering model according to the size relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value may include:

[0012] When the wake-up trigger value is less than or equal to the wake-up score, determining a first preset value as the reduction factor;

[0013] When the wake-up mean value is less than or equal to the wake-up score, and the wake-up score is less than the wake-up trigger value, determining a second preset value as the reduction factor;

[0014] When the wake-up score is less than the wake-up mean value, determining a third preset value as the reduction factor;

[0015] The first preset value is smaller than the second preset value, and the second preset value is smaller than the third preset value.

[0016] In some embodiments, before obtaining the wake-up score of the first voice frame signal of the wake-up word, the method further includes:

[0017] For each frame of the historical speech frame signal of the wake-up word in the historical wake-up speech set, obtain a target historical wake-up score of the target historical speech frame signal containing the wake-up word;

[0018] The mean of the target historical wake-up distribution is calculated to obtain the wake-up mean.

[0019] In some embodiments, after updating the filtering model according to the update step size and the reduction factor, the method further includes:

[0020] Filter the second voice frame signal of the wake-up word according to the updated filtering model to obtain a filtered second voice frame signal, wherein the second voice frame signal of the wake-up word is the next frame signal of the first voice frame signal of the wake-up word;

[0021] The filtered second voice frame signal is sent to a wake-up device, so that the wake-up device wakes up the target device according to the magnitude relationship between the wake-up score of the filtered second voice frame signal and the wake-up trigger threshold.

[0022] In some embodiments, the method further comprises:

[0023] When the wake-up score is greater than or equal to the wake-up trigger threshold, the filtering model is not updated.

[0024] In a second aspect, an embodiment of the present application provides a filtering model updating device, the device comprising:

[0025] an acquisition module, configured to acquire a wake-up score of a first voice frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first voice frame signal contains the wake-up word;

[0026] A calculation module, configured to calculate a wake-up trigger value according to a wake-up trigger threshold and a wake-up score mean, wherein the wake-up score mean is a mean value calculated according to the wake-up score distribution of a historical speech frame signal of a wake-up word in a historical wake-up speech set;

[0027] a determination module, configured to determine, when the wake-up score is less than the wake-up trigger threshold, a reduction factor of the update step size of the filtering model according to a relationship among the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value;

[0028] An updating module is used to update the filtering model according to the update step size and the reduction factor.

[0029] In some embodiments, the determining module may include:

[0030] A first determining submodule, configured to determine a first preset value as the reduction factor when the wake-up trigger value is less than or equal to the wake-up score;

[0031] A second determining submodule, configured to determine a second preset value as the reduction factor when the wake-up mean value is less than or equal to the wake-up score and the wake-up score is less than the wake-up trigger value;

[0032] A third determining submodule, configured to determine a third preset value as the reduction factor when the wake-up score is less than the wake-up mean value;

[0033] The first preset value is smaller than the second preset value, and the second preset value is smaller than the third preset value.

[0034] In a third aspect, an embodiment of the present application provides a filtering model updating device, the device comprising: a processor and a memory storing computer program instructions;

[0035] When the processor executes the computer program instructions, the filtering model updating method described in any embodiment of the present application is implemented.

[0036] In a fourth aspect, an embodiment of the present application provides a computer storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, a filtering model updating method as described in any embodiment of the present application is implemented.

[0037] In a fifth aspect, an embodiment of the present application provides a computer program product. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes a filtering model updating method as described in any embodiment of the present application.

[0038] The filter model updating method, device, equipment, medium and computer program product of the embodiment of the present application calculate the wake-up trigger value according to the wake-up trigger threshold and the wake-up score mean. Then, according to the size relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value, the reduction factor of the update step of the filter model is determined, and finally the filter model is updated according to the update step and the reduction factor. The wake-up score mean is the mean calculated according to the wake-up score distribution of the historical voice frame signal of the wake-up word in the historical wake-up voice set, which characterizes the average possibility that the voice frame signal of the wake-up word can wake up the device. The wake-up trigger threshold indicates that the voice frame signal of the wake-up word can wake up the device. The wake-up trigger value calculated by combining the wake-up trigger threshold and the wake-up score mean can represent the average proximity of the voice frame signal of the wake-up word to the event that can wake up the device. In this way, the reduction factor is a factor that comprehensively considers the possibility of the first voice frame signal of the wake-up word containing the wake-up word, the proximity to the event that can wake up the device, the average possibility that the voice frame signal of the wake-up word can wake up the device, and the wake-up trigger threshold. In other words, the reduction factor is a factor determined according to the different proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. Then, the filtering model is updated according to the update step size and the reduction factor, which takes into account the different degrees of proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. The filtering model is updated to different degrees, effectively avoiding the erroneous update of the filtering model to the voice frame signal segment containing the wake-up word, reducing the damage of the filter coefficient update to the voice frame signal, and improving the accuracy of the filtering model update. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0040] Figure 1 It is a flowchart of a filtering model updating method provided by an embodiment of the present application;

[0041] Figure 2 It is a flowchart of another filtering model updating method provided in an embodiment of the present application;

[0042] Figure 3 This is a flowchart of another filtering model updating method provided in an embodiment of the present application;

[0043] Figure 4 This is a flowchart of another filtering model updating method provided in an embodiment of the present application;

[0044] Figure 5 is a schematic diagram of a filter model updating device provided in an embodiment of the present application;

[0045] Figure 6 It is a schematic diagram of a filtering model updating device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0047] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0048] Before explaining the technical solutions provided by the embodiments of the present application, in order to facilitate the understanding of the embodiments of the present application, the present application first introduces the special terms.

[0049] Wake-up trigger threshold: During voice wake-up, the device wake-up is triggered when the wake-up word is detected. The wake-up word is distributed in several frames of voice frame signals, which means that the voice frame signal corresponding to the wake-up word will last for a period of time. The wake-up trigger is completed in an instant, but during the entire duration of the wake-up word, each voice frame signal will have a corresponding possibility of containing the wake-up word. The accumulation of the possibility of voice frame signals containing the wake-up word over a period of time can form an event to wake up the device. The value indicating the occurrence of this event is the wake-up trigger threshold.

[0050] As an example, assuming that the duration of a wake-up word is f frames, it will gradually accumulate to a value F during the entire period. When F is greater than or equal to the wake-up trigger threshold T, the device is woken up; if F is less than T, the device is not woken up.

[0051] The current filter model will filter out invalid and interfering speech such as noise on the current voice frame signal, and will automatically update and iterate while filtering the voice signal. Generally speaking, the update of the filter model is based on whether the device is awakened, that is, if it is not awakened, it will be updated, and if it is awakened, it will not be updated.

[0052] However, the inventors have found that this kind of update based on whether the device is awakened or not will result in the filter model being updated when the current voice frame signal filtered by the filter model is a voice frame signal containing a wake-up word, but the device has not yet been awakened. In other words, a "one-size-fits-all" update is performed, which will cause greater damage to the filtering of the voice frame signal containing the wake-up word.

[0053] Therefore, the inventor proposed to calculate the wake-up trigger value based on the wake-up trigger threshold and the wake-up score mean. Then, based on the relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value, determine the reduction factor of the update step of the filter model, and finally update the filter model based on the update step and the reduction factor. After considering the different degrees of proximity of the current voice frame signal of the wake-up word to the event that can wake up the device, the filter model is updated to different degrees to avoid a "one-size-fits-all" update of the filter model and improve the accuracy of the filter model update.

[0054] The filtering model updating method of the present application can be executed by a filtering model updating device.

[0055] The following describes an embodiment of the present application in conjunction with the accompanying drawings. First, a filtering model updating method provided in an embodiment of the present application is introduced.

[0056] Figure 1 A schematic flow chart of a filtering model updating method provided in an embodiment of the present application is shown, and the method shown includes:

[0057] S110, obtaining a wake-up score of a first voice frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first voice frame signal contains the wake-up word.

[0058] S120, calculating a wake-up trigger value according to the wake-up trigger threshold and the wake-up score mean, wherein the wake-up score mean is a mean value calculated according to the wake-up score distribution of the historical speech frame signals of the wake-up words in the historical wake-up speech set.

[0059] S130: When the wake-up score is less than the wake-up trigger threshold, determine a reduction factor of the update step size of the filtering model according to a relationship among the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value.

[0060] S140: Update the filtering model according to the update step size and the reduction factor.

[0061] In the embodiment of the present application, the wake-up trigger value is calculated according to the wake-up trigger threshold and the wake-up score mean. Then, according to the size relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value, the reduction factor of the update step of the filter model is determined, and finally, the filter model is updated according to the update step and the reduction factor. The wake-up score mean is the mean calculated according to the wake-up score distribution of the historical voice frame signal of the wake-up word in the historical wake-up voice set, which characterizes the average possibility that the voice frame signal of the wake-up word can wake up the device. The wake-up trigger threshold indicates that the voice frame signal of the wake-up word can wake up the device. The wake-up trigger value calculated by combining the wake-up trigger threshold and the wake-up score mean can represent the average proximity of the voice frame signal of the wake-up word to the event that can wake up the device. In this way, the reduction factor is a factor that comprehensively considers the possibility of the first voice frame signal of the wake-up word containing the wake-up word, the proximity to the event that can wake up the device, the average possibility that the voice frame signal of the wake-up word can wake up the device, and the wake-up trigger threshold. In other words, the reduction factor is a factor determined according to the different proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. Then, the filtering model is updated according to the update step size and the reduction factor, which takes into account the different degrees of proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. The filtering model is updated to different degrees, effectively avoiding the erroneous update of the filtering model to the voice frame signal segment containing the wake-up word, reducing the damage of the filter coefficient update to the voice frame signal, and improving the accuracy of the filtering model update.

[0062] In some embodiments, in S110, each frame of the speech frame signal of the wake-up word will correspond to a wake-up score. The first speech frame signal can be the current speech frame signal of the wake-up word. The wake-up score corresponding to the current speech frame signal is a cumulative score calculated by a wake-up device independent of the filter model updating device in combination with the wake-up score of the speech frame signal of the previous frame and the possibility of the current speech frame signal itself containing the wake-up word. The filter model updating device receives the wake-up score of the current speech frame signal calculated by the wake-up device. The embodiment of the present application does not limit the accumulation method, which can be any mathematical calculation method.

[0063] The wake-up score is used to indicate the possibility that the current speech frame signal contains the wake-up word. The wake-up score is a score related to the possibility that the speech frame signal of the previous frame contains the wake-up word, and is an accumulated score.

[0064] As an example, there are 1000 frames of speech frame signals of the wake-up word, among which the 200th frame to the 900th frame contain the wake-up word. The current speech frame signal is the 700th frame, and the probabilities of the speech frame signals of the 1st to 699th frames containing the wake-up word correspond to j_1 to j_699 respectively, and the probability of the speech frame signal of the 700th frame containing the wake-up word is j_700. Then, the wake-up score of the 700th frame can be the result obtained by any data calculation method containing all values ​​from j_1 to j_700. For example, the wake-up score of the 700th frame can be j_1+…+j_699+j_700.

[0065] In some embodiments, the wake-up score of the speech frame signal of the wake-up word is calculated in different ways for different speech wake-up technologies, and the specific calculation method is a prior art, which is not described in detail in the embodiments of this application. Among them, speech wake-up technologies include but are not limited to confidence-based speech wake-up, recognition engine-based speech wake-up, filler-based speech wake-up, and deep learning-based speech wake-up.

[0066] In order to improve the accuracy of updating the filtering model, in some embodiments, Figure 2 The flowchart of another filtering model updating method provided by an embodiment of the present application shown in FIG. 1 further includes S210-S220 before obtaining the wake-up score of the first voice frame signal of the wake-up word:

[0067] S210, for each frame of the historical speech frame signal of the wake-up word in the historical wake-up speech set, obtain a target historical wake-up score of the target historical speech frame signal containing the wake-up word.

[0068] In some embodiments, in S210, the target historical wake-up score is the wake-up score of each frame of the speech frame signal containing the wake-up word in the historical wake-up speech set. The filter model updating device receives the wake-up score of each frame of the historical speech frame signal containing the wake-up word in the historical wake-up speech set calculated by the wake-up device.

[0069] S220, counting the mean of the target historical wake-up distribution to obtain the wake-up mean.

[0070] In some embodiments, in S220, when the historical wake-up speech set is not updated, the wake-up mean used by the filtering model updating device in the future is the mean of the target historical wake-up distribution previously counted by the filtering model updating device, and this mean is the wake-up mean; when the historical wake-up speech set is updated, the filtering model updating device repeats steps 210-220 to re-count the mean of the target historical wake-up distribution.

[0071] In some embodiments, the embodiments of the present application do not limit the specific method for the filter model update device to calculate the mean of the target historical wake-up distribution. The wake-up distribution of the speech frame signal containing the wake-up word in each frame in different wake-up words and different historical wake-up speech sets is different, and the method of calculating the mean is also different.

[0072] As an example, the filtering model updating device may calculate the mean of the statistical target historical wake-up score distribution by calculating the weighted average of the wake-up scores of each speech frame signal containing the wake-up word in the historical wake-up speech set, and use the weighted average as the wake-up mean.

[0073] In the embodiment of the present application, the wake-up mean is the mean obtained by statistically analyzing the target historical wake-up points of the target historical voice frame signal containing the wake-up word in each frame of the historical wake-up voice set. It is the mean after considering the probability value distribution of most historical voice frame signals containing the wake-up word, and more accurately represents the average possibility of the voice frame signal of the wake-up word waking up the device. The corresponding wake-up mean can be obtained for different historical voice sets and different wake-up words. In this way, the wake-up mean that can adapt to different wake-up words and historical wake-up voice sets, and represents the average possibility of the voice frame signal of the wake-up word waking up the device, can determine the wake-up trigger value more accurately, and further, according to the wake-up trigger value, the wake-up mean, the wake-up trigger threshold and the wake-up point, it is possible to more accurately judge the different degrees of the first voice frame signal of the wake-up word from the event of waking up the device, thereby improving the accuracy of determining the reduction factor, and further improving the accuracy of updating the filter model.

[0074] In some embodiments, in S120, the filter model updating device calculates a value that can represent the average proximity of the voice frame signal of the wake-up word to the event of waking up the device based on the wake-up trigger threshold and the wake-up score mean. This value is called the wake-up trigger value. The embodiments of the present application do not limit the specific calculation of the wake-up trigger value.

[0075] As an example, the filter model updating device may calculate the wake-up trigger value by using Formula 1, wherein T is the wake-up trigger threshold, t is the wake-up trigger value, m is the wake-up score average, and r is the proportional coefficient.

[0076] t=m+r*(Tm) Formula 1

[0077] The value of r is adjusted according to different voice wake-up technologies and actual conditions, and is not specifically limited in the embodiments of the present application.

[0078] In some embodiments, the value of r can be determined by the degree of difference between the wake-up score of the historical speech frame signal containing the wake-up word in the historical wake-up speech set and the mean wake-up score. For example, the filter model update device calculates the sum of the wake-up scores of each historical speech frame signal containing the wake-up word in the historical wake-up speech set, and then calculates the difference between it and the mean wake-up score. When the difference is greater than or equal to the preset threshold, r takes the first preset value; when the difference is less than the preset threshold, r takes the second preset value, and the first preset value is greater than the second preset value. The embodiment of the present application does not limit the values ​​of the preset threshold and the first threshold and the second threshold, and is adjusted according to different voice wake-up technologies and actual conditions.

[0079] In order to improve the accuracy of determining the reduction factor, in some embodiments, the value of r may be 0.8.

[0080] In some embodiments, the wake-up trigger threshold is a numerical value calculated based on the historical voice frame signal of the wake-up word in the historical wake-up voice set, which is used to indicate that the voice frame signal of the wake-up word can wake up the device. There are different calculation methods for calculating the wake-up trigger threshold for different voice wake-up technologies, but the specific calculation method is a prior art and is not described in detail in the embodiments of this application. Among them, voice wake-up technologies include but are not limited to confidence-based voice wake-up, recognition engine-based voice wake-up, junk word-based voice wake-up, deep learning-based voice wake-up, and neural network-based voice wake-up. For example, taking deep learning-based voice wake-up as an example, the wake-up trigger threshold can be the confidence obtained after counting the posterior probability of each historical voice frame signal containing the wake-up word in the historical wake-up voice set output by the neural network.

[0081] In some embodiments, in S130, when the wake-up score of the current voice frame signal is less than the wake-up trigger threshold, the filtering model updating device uses the different size relationships among the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value to represent the different degrees of proximity of the current voice frame signal of the wake-up word to the event of waking up the device, and then determines the reduction factors of different update steps.

[0082] In some embodiments, the reduction factor has a value range of (0, 1].

[0083] In some embodiments, the filtering model may include an adaptive filter, and the updating of the filter coefficient of the adaptive filter is related to the filter coefficient and the update step size of the previous frame. The adaptive filter update step size has different calculation methods for different adaptive filtering algorithms, and its specific calculation method is a prior art, and the embodiments of the present application are not repeated. Adaptive filtering algorithms include but are not limited to the least mean square adaptive filtering (Least Mean Square, LMS) algorithm, the normalized least mean square adaptive filtering (Normalized Least Mean Square, NLMS) algorithm, the neighbor propagation (Affinity Propagation, AP) algorithm, the derivative algorithm of the LMS algorithm, the derivative algorithm of the NLMS algorithm, and the partitioned block frequency domain adaptive filtering algorithm (Partitioned Block Frequency Domain Adaptive Filter, PBFDAF).

[0084] In order to improve the accuracy of updating the filtering model, in some embodiments, when the wake-up score is less than the wake-up trigger threshold, determining the reduction factor of the update step size of the filtering model according to the relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value may include:

[0085] When the wake-up trigger value is less than or equal to the wake-up score, a first preset value is determined as the reduction factor.

[0086] In some embodiments, continuing to refer to the above embodiments, since the calculation of the wake-up trigger value is related to the wake-up trigger threshold, when the wake-up trigger value is less than or equal to the wake-up score, it can be considered that the current voice frame signal of the wake-up word is very close to the event that can wake up the device, and it can be considered that the wake-up score of the current voice frame signal of the wake-up word is relatively close to the wake-up trigger threshold and almost triggers the wake-up. In other words, the current voice frame signal of the wake-up word is very likely to contain the wake-up word, that is, it is very likely to be in the wake-up word segment in the entire wake-up word voice signal. At this time, the filter model update device uses a smaller first preset value as the reduction factor.

[0087] In some embodiments, in order to more accurately control the update of the filtering model, the first preset value may be 0.1.

[0088] It is worth noting that in the embodiment of the present application, there is no specific limitation on the value of the first preset value, and it can be adjusted in response to actual conditions.

[0089] When the wake-up mean value is less than or equal to the wake-up score, and the wake-up score is less than the wake-up trigger value, a second preset value is determined as the reduction factor.

[0090] In some embodiments, continuing to refer to the above embodiments, when the wake-up mean is less than or equal to the wake-up score, and the wake-up score is less than the wake-up trigger value, it can be considered that the current voice frame signal of the wake-up word is not as close to the event of being able to wake up the device as when the wake-up trigger value is less than or equal to the wake-up score, but the wake-up score of the current voice frame signal of the wake-up word is closer to the wake-up trigger value, that is, closer to the average closeness to the event of being able to wake up the device. It can be considered that at this time, the possibility of the current voice frame signal of the wake-up word waking up the device is in a medium state. In other words, the current voice frame signal of the wake-up word may be close to the voice frame signal containing the wake-up word, and is in the front section of the wake-up word segment in the entire voice signal of the wake-up word. At this time, the filtering model updating device uses a second preset value that is greater than the first preset value as a reduction factor.

[0091] In some embodiments, in order to more accurately control the update of the filtering model, the second preset value may be 0.5.

[0092] It is worth noting that in the embodiment of the present application, there is no specific limitation on the value of the second preset value, and it can be adjusted in response to actual conditions.

[0093] When the wake-up score is less than the wake-up mean value, a third preset value is determined as the reduction factor.

[0094] The first preset value is smaller than the second preset value, and the second preset value is smaller than the third preset value.

[0095] In some embodiments, referring to the above embodiment, when the wake-up score is less than the wake-up mean, that is, the wake-up score of the current voice frame signal of the wake-up word is less than the average possibility of the voice frame signal of the wake-up word waking up the device. Basically, the device will not be woken up. At this time, the filter model updating device uses a third preset value greater than the second preset value as a reduction factor.

[0096] In some embodiments, in order to more accurately control the update of the filtering model, the third preset value may be 1.

[0097] It is worth noting that in the embodiment of the present application, there is no specific limitation on the value of the first preset value, and it can be adjusted in response to actual conditions.

[0098] In the embodiment of the present application, the reduction factor is determined based on the different size relationships among the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value, taking into account the different proximity of the voice frame signal of the wake-up word to the event that can wake up the device. The first preset value is less than the second preset value, and the second preset value is less than the third preset value, which means that the reduction factor is reduced to different degrees according to the different proximity of the voice frame signal of the wake-up word to the event that can wake up the device, and thus the update strength of the filter model can be controlled, thereby achieving different update strengths of the filter model according to the different proximity of the voice frame signal of the wake-up word to the event that can wake up the device, thereby improving the accuracy of updating the filter model.

[0099] In some embodiments, in S140, the filter model updating device updates the filter coefficients of the filter model according to the reduced update step size obtained by multiplying the update step size by the reduction factor to implement the update of the filter model.

[0100] As an example, using an adaptive filter and an LMS algorithm, updating the filter model is to update the filter coefficients of the adaptive filter, which can be specifically shown in Formula 2.

[0101] W(n+1)=W(n)+k*μ*X(n)*e(n) Formula 2

[0102] Wherein, W(n+1) is the updated filter coefficient, W(n) is the filter coefficient when filtering the current speech frame signal, k is the reduction factor, μ is the update step size, X(n) is the current input noise signal, and e(n) is the instantaneous input error power.

[0103] In order to improve the accuracy of updating the filtering model, in some embodiments, Figure 3 As shown, a flowchart of a filtering model updating method provided by an embodiment of the present application, the method further includes S310:

[0104] S110, obtaining a wake-up score of a first voice frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first voice frame signal contains the wake-up word.

[0105] S120, calculating a wake-up trigger value according to the wake-up trigger threshold and the wake-up score mean, wherein the wake-up score mean is a mean value calculated according to the wake-up score distribution of the historical speech frame signals of the wake-up words in the historical wake-up speech set.

[0106] S130: When the wake-up score is less than the wake-up trigger threshold, determine a reduction factor of the update step size of the filtering model according to a relationship among the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value.

[0107] S140: Update the filtering model according to the update step size and the reduction factor.

[0108] S310: When the wake-up score is greater than or equal to the wake-up trigger threshold, the filtering model is not updated.

[0109] S110 to S140 are the same as S110 to S140 in the above embodiment, and for the sake of brevity, they are not described in detail here.

[0110] In some actual examples, in S310, when the wake-up score of the current voice frame signal of the wake-up word is greater than or equal to the wake-up trigger threshold, it means that the accumulated wake-up score of the current voice frame signal in the entire voice signal of the wake-up word indicates that there is a wake-up word that can wake up the device. In other words, the filter model that filters the current voice frame signal can just make the wake-up word in the voice frame signal, and filter out invalid and interfering signals. The filter model update device receives an identification signal that the filter model that filters the current voice frame signal has been updated and does not need to be updated. In response to the identification signal, the filter model update device no longer updates the filter coefficients of the filter model.

[0111] In some embodiments, when using the filter model of the previous frames to filter the speech frame signal of the current frame, the filter model update device will determine whether the filter coefficient of the current filter model is reasonable after receiving the identification signal. If the current filter coefficient is the filter coefficient updated after filtering the speech frame signal containing the wake-up word, the filter model update device will invalidate the current filter coefficient according to the length d of the wake-up word in the received identification signal, and use the filter coefficient d frames away from the current filter coefficient as the last updated filter coefficient.

[0112] In the embodiment of the present application, when the wake-up score is equal to the wake-up trigger threshold, the reduction factor is determined according to the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value, and then the filter model is updated according to the update step size and the reduction factor, so as to control the update intensity of the filter model according to the different proximity of the first voice frame signal of the wake-up word to the event of waking up the device. When the wake-up score is greater than or equal to the wake-up trigger threshold, the filter model is no longer updated, taking into account the different situations in which the voice frame signal of the wake-up word wakes up the device and the voice frame signal of the wake-up word cannot wake up the device, thereby improving the accuracy of updating the filter model.

[0113] In the existing scheme, when filtering invalid signals and interference signals such as noise in the voice frame signal of the wake-up word, a delay filter algorithm is used to filter the voice frame signal in order to improve the recognition rate of the wake-up algorithm. Generally speaking, an adaptive filter is used to select two signals from the microphone array for adaptive filtering, and the updated filter coefficient h(x) is delayed. The delay time depends on the empirical statistics of the length of the wake-up word. Assuming that the length of the wake-up word is d, the filter model corresponding to the filter coefficient h(xd) of the xdth frame is used to filter the xth frame voice frame signal, and the filter coefficient is no longer updated when the device is woken up.

[0114] However, the inventors have found that in the existing scheme, although the filter model corresponding to the filter coefficient of the previous xd frames is used to filter the xth frame voice frame signal, it is ensured that when the voice frame signal reaches the end of the wake-up word, the filter coefficient used is not affected by the wake-up word, and thus no signal elimination occurs. However, in actual application scenarios, since the update of the filter coefficient and the wake-up device are independently completed in the filter model update device and the wake-up device, respectively, when the filter model update device updates the filter model, the current voice frame signal fails to trigger the wake-up in the wake-up device, and the current voice frame signal may be a voice frame signal containing the wake-up word, but the filter model update device at this time will still update the filter model. In other words, in the prior art solution, the filter model update device will still update the wake-up word segment in the entire voice signal containing the wake-up word. Then, when continuing to wake up or recognize, there will be strong damage to the voice frame signal, thereby affecting the recognition rate of the voice frame signal, and ultimately resulting in a poor interactive experience.

[0115] Therefore, the inventor proposes that in order to improve the recognition rate of the voice frame signal of the wake-up word, in some embodiments, Figure 4 As shown, a flowchart of another method for updating a filter model provided by an embodiment of the present application is shown. After updating the filter model according to the update step size and the reduction factor, it also includes S410-S420:

[0116] S410, filtering the second voice frame signal of the wake-up word according to the updated filtering model to obtain a filtered second voice frame signal, wherein the second voice frame signal of the wake-up word is the next frame signal of the first voice frame signal of the wake-up word.

[0117] In some embodiments, in S410, the filter model updating device filters the second speech frame signal of the wake-up word through the updated filter model.

[0118] As an example, it can be understood that when the filtering model corresponding to the filtering coefficients of the previous xd frames is used to filter the x-th frame speech frame signal, the filtering model updated by the filtering model updating device is the filtering model corresponding to the x-d+1 frame, thereby completing the filtering of the x+1-th frame speech frame signal.

[0119] S420, sending the filtered second voice frame signal to the wake-up device, so that the wake-up device sends the filtered second voice frame signal to the wake-up device, so that the wake-up device wakes up the target device according to the relationship between the wake-up score of the filtered second voice frame signal and the wake-up trigger threshold.

[0120] In some implementations, in S420, the filter model updating device sends the filtered second voice frame signal to the wake-up device, and the wake-up device calculates the wake-up score of the filtered second voice frame signal and compares the wake-up score with the wake-up trigger threshold. When the wake-up score is greater than or equal to the wake-up trigger threshold, the wake-up device wakes up the device corresponding to the wake-up word; when the wake-up score is lower than the wake-up trigger threshold, the wake-up device does not wake up the device corresponding to the wake-up word.

[0121] In an embodiment of the present application, the next frame signal of the first voice frame signal of the wake-up word, that is, the second voice frame signal of the wake-up word, is filtered with an updated filtering model. The wake-up device wakes up the target device according to the relationship between the wake-up score of the filtered second voice frame signal and the wake-up trigger threshold. The updated filtering model is a filtering model that takes into account the different degrees of proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. If it is effective, the filtering damage to the second voice frame signal will be smaller, thereby improving the recognition rate of the voice frame signal and further increasing the user's interactive experience.

[0122] Based on the filtering model updating method provided in any of the above embodiments, the present application also provides a filtering model updating device embodiment. Figure 6

[0123] Figure 5 FIG. 2 shows a schematic diagram of a filter model updating device provided by an embodiment of the present application. Figure 5 As shown, the device may include:

[0124] An acquisition module 510 is used to acquire a wake-up score of a first speech frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first speech frame signal contains the wake-up word;

[0125] A calculation module 520, configured to calculate a wake-up trigger value according to a wake-up trigger threshold and a wake-up score mean, wherein the wake-up score mean is a mean value calculated according to the wake-up score distribution of the historical speech frame signal of the wake-up word in the historical wake-up speech set;

[0126] A determination module 530, configured to determine a reduction factor of an update step length of a filtering model according to a relationship among the wakeup trigger threshold, the wakeup mean, the wakeup score, and the wakeup trigger value when the wakeup score is less than the wakeup trigger threshold;

[0127] The updating module 540 is used to update the filtering model according to the updating step size and the reduction factor.

[0128] The device in the embodiment of the present application calculates the wake-up trigger value according to the wake-up trigger threshold and the wake-up score mean. Then, according to the size relationship between the wake-up trigger threshold, the wake-up mean, the wake-up score and the wake-up trigger value, the reduction factor of the update step of the filter model is determined, and finally the filter model is updated according to the update step and the reduction factor. The wake-up score mean is the mean calculated according to the wake-up score distribution of the historical voice frame signal of the wake-up word in the historical wake-up voice set, which characterizes the average possibility that the voice frame signal of the wake-up word can wake up the device. The wake-up trigger threshold indicates that the voice frame signal of the wake-up word can wake up the device. The wake-up trigger value calculated by combining the wake-up trigger threshold and the wake-up score mean can represent the average proximity of the voice frame signal of the wake-up word to the event that can wake up the device. In this way, the reduction factor is a factor that comprehensively considers the possibility of the first voice frame signal of the wake-up word containing the wake-up word, the proximity to the event that can wake up the device, the average possibility that the voice frame signal of the wake-up word can wake up the device, and the wake-up trigger threshold. In other words, the reduction factor is a factor determined according to the different proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. Then, the filtering model is updated according to the update step size and the reduction factor, which takes into account the different degrees of proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. The filtering model is updated to different degrees, effectively avoiding the erroneous update of the filtering model to the voice frame signal segment containing the wake-up word, reducing the damage of the filter coefficient update to the voice frame signal, and improving the accuracy of the filtering model update.

[0129] In some embodiments, in order to improve the accuracy of updating the filtering model, the determination module 530 may include:

[0130] A first determining submodule, configured to determine a first preset value as the reduction factor when the wake-up trigger value is less than or equal to the wake-up score;

[0131] A second determining submodule, configured to determine a second preset value as the reduction factor when the wake-up mean value is less than or equal to the wake-up score and the wake-up score is less than the wake-up trigger value;

[0132] A third determining submodule, configured to determine a third preset value as the reduction factor when the wake-up score is less than the wake-up mean value;

[0133] The first preset value is smaller than the second preset value, and the second preset value is smaller than the third preset value.

[0134] In the device in the embodiment of the present application, the reduction factor is determined based on the different size relationships between the wake-up trigger threshold, the wake-up mean, the wake-up score, and the wake-up trigger value, taking into account the different proximity of the voice frame signal of the wake-up word to the event that can wake up the device. The first preset value is less than the second preset value, and the second preset value is less than the third preset value, which means that the reduction factor is reduced to different degrees according to the different proximity of the voice frame signal of the wake-up word to the event that can wake up the device, and thus the update strength of the filter model can be controlled, thereby achieving different update strengths of the filter model according to the different proximity of the voice frame signal of the wake-up word to the event that can wake up the device, thereby improving the accuracy of updating the filter model.

[0135] In some embodiments, in order to improve the accuracy of updating the filter model, before obtaining the wake-up score of the first voice frame signal of the wake-up word, the acquisition module 510 further includes:

[0136] The acquisition submodule is used to obtain the target historical wake-up score of the target historical speech frame signal containing the wake-up word for each frame of the historical wake-up speech set.

[0137] The statistical submodule is used to count the mean of the target historical wake-up distribution to obtain the wake-up mean.

[0138] In the device in the embodiment of the present application, the wake-up mean is the mean obtained by statistically analyzing the target historical wake-up points of the target historical voice frame signal containing the wake-up word in each frame of the historical wake-up voice set. It is the mean after considering the probability value distribution of most historical voice frame signals containing the wake-up word, and more accurately represents the average possibility of the voice frame signal of the wake-up word waking up the device. The corresponding wake-up mean can be obtained for different historical voice sets and different wake-up words. In this way, the wake-up mean that can adapt to different wake-up words and historical wake-up voice sets, and represents the average possibility of the voice frame signal of the wake-up word waking up the device, can determine the wake-up trigger value more accurately, and further, according to the wake-up trigger value, the wake-up mean, the wake-up trigger threshold and the wake-up point, it is possible to more accurately judge the different degrees of the first voice frame signal of the wake-up word from the event of waking up the device, thereby improving the accuracy of determining the reduction factor, and further improving the accuracy of updating the filter model.

[0139] In some embodiments, in order to improve the recognition rate of the voice frame signal of the wake-up word, after updating the filtering model according to the update step size and the reduction factor, the updating module 540 further includes:

[0140] The filtering submodule is used to filter the second speech frame signal of the wake-up word according to the updated filtering model to obtain a filtered second speech frame signal, wherein the second speech frame signal of the wake-up word is the next frame signal of the first speech frame signal of the wake-up word.

[0141] The sending submodule is used to send the filtered second voice frame signal to the wake-up device, so that the wake-up device sends the filtered second voice frame signal to the wake-up device, so that the wake-up device wakes up the target device according to the size relationship between the wake-up score of the filtered second voice frame signal and the wake-up trigger threshold.

[0142] The device in the embodiment of the present application uses an updated filtering model to filter the next frame signal of the first voice frame signal of the wake-up word, that is, the second voice frame signal of the wake-up word. The wake-up device wakes up the target device according to the relationship between the wake-up score of the filtered second voice frame signal and the wake-up trigger threshold. The updated filtering model is a filtering model that takes into account the different degrees of proximity of the first voice frame signal of the wake-up word to the event that can wake up the device. If it is effective, the filtering damage to the second voice frame signal will be smaller, thereby improving the recognition rate of the voice frame signal and further increasing the user's interactive experience.

[0143] In addition, combined with the filter model updating method in the above embodiment, Figure 6 As shown, an embodiment of the present application may provide a filtering model updating device, which may include a processor 610 and a memory 620 storing computer program instructions.

[0144] Specifically, the processor 610 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0145] The memory 620 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 620 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 620 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 620 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 620 is a non-volatile solid-state memory. In a specific embodiment, the memory 620 includes a read-only memory (ROM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM) or a flash memory or a combination of two or more of these.

[0146] The processor 610 implements any one of the filtering model updating methods in the above embodiments by reading and executing computer program instructions stored in the memory 620 .

[0147] In one example, the electronic device may further include a communication interface 630 and a bus 640. Figure 6 As shown, the processor 610, the memory 620, and the communication interface 630 are connected via a bus 640 and communicate with each other.

[0148] The communication interface 630 is mainly used to implement communication between various modules, devices, units and / or devices in the embodiments of the present application.

[0149] Bus 640 includes hardware, software or both, and the parts of electronic equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 640 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0150] The filter model updating processor implements the filter model updating method described in any of the above embodiments when executing the computer program instructions.

[0151] In addition, in combination with the above-mentioned filtering model updating method, an embodiment of the present application may provide a computer storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the filtering model updating method described in any of the above-mentioned embodiments is implemented.

[0152] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0153] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0154] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0155] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0156] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A filtering model updating method, characterized in that: include: Obtaining a wake-up score of a first speech frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first speech frame signal contains the wake-up word; Calculate the wake-up trigger value according to the wake-up trigger threshold and the wake-up score mean, wherein the wake-up score mean is the mean value calculated according to the wake-up score distribution of the historical speech frame signal of the wake-up word in the historical wake-up speech set; When the wake-up score is less than the wake-up trigger threshold, determining a reduction factor of the update step size of the filtering model according to the relationship among the wake-up trigger threshold, the wake-up score average, the wake-up score, and the wake-up trigger value; updating the filtering model according to the update step size and the reduction factor; When the wake-up score is less than the wake-up trigger threshold, determining the reduction factor of the update step size of the filtering model according to the relationship between the wake-up trigger threshold, the wake-up score average, the wake-up score, and the wake-up trigger value, specifically includes: When the wake-up trigger value is less than or equal to the wake-up score, determining a first preset value as the reduction factor; When the wake-up score average is less than or equal to the wake-up score, and the wake-up score is less than the wake-up trigger value, determining a second preset value as the reduction factor; When the wake-up score is less than the wake-up score average, determining a third preset value as the reduction factor; The first preset value is smaller than the second preset value, and the second preset value is smaller than the third preset value.

2. The method according to claim 1, characterized in that Before obtaining the wake-up score of the first voice frame signal of the wake-up word, the method further includes: For each frame of the historical speech frame signal of the wake-up word in the historical wake-up speech set, obtain a target historical wake-up score of the target historical speech frame signal containing the wake-up word; The mean of the target historical wake-up score distribution is calculated to obtain the wake-up score mean.

3. The method according to claim 1, characterized in that After updating the filtering model according to the update step size and the reduction factor, the method further includes: Filter the second voice frame signal of the wake-up word according to the updated filtering model to obtain a filtered second voice frame signal, wherein the second voice frame signal of the wake-up word is the next frame signal of the first voice frame signal of the wake-up word; The filtered second voice frame signal is sent to a wake-up device, so that the wake-up device wakes up the target device according to the magnitude relationship between the wake-up score of the filtered second voice frame signal and the wake-up trigger threshold.

4. The method according to claim 1, characterized in that: The method further comprises: When the wake-up score is greater than or equal to the wake-up trigger threshold, the filtering model is not updated.

5. A filter model updating device, characterized in that: The device comprises: an acquisition module, configured to acquire a wake-up score of a first voice frame signal of a wake-up word, wherein the wake-up score is used to indicate the possibility that the first voice frame signal contains the wake-up word; A calculation module, configured to calculate a wake-up trigger value according to a wake-up trigger threshold and a wake-up score mean, wherein the wake-up score mean is a mean value calculated according to the wake-up score distribution of a historical speech frame signal of a wake-up word in a historical wake-up speech set; a determination module, configured to determine, when the wake-up score is less than the wake-up trigger threshold, a reduction factor of the update step size of the filtering model according to the relationship between the wake-up trigger threshold, the wake-up score average, the wake-up score, and the wake-up trigger value; An updating module, used for updating the filtering model according to the updating step size and the reduction factor; Determine the modules, including: A first determining submodule, configured to determine a first preset value as the reduction factor when the wake-up trigger value is less than or equal to the wake-up score; A second determining submodule, configured to determine a second preset value as the reduction factor when the wake-up score average is less than or equal to the wake-up score and the wake-up score is less than the wake-up trigger value; A third determining submodule, configured to determine a third preset value as the reduction factor when the wake-up score is less than the wake-up score average; The first preset value is smaller than the second preset value, and the second preset value is smaller than the third preset value.

6. A filter model updating device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the filtering model updating method according to any one of claims 1 to 4 is implemented.

7. A computer storage medium, characterized in that: The computer storage medium stores computer program instructions, and when the computer program instructions are executed by the processor, the filtering model updating method according to any one of claims 1 to 4 is implemented.

8. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the filtering model updating method as described in any one of claims 1 to 4.

Images