Noise reduction parameter optimization method, device, terminal equipment and medium

Abstract

The embodiments of the present application propose a noise reduction parameter optimization method, apparatus, terminal device and medium, wherein, in the above-mentioned optimization method, the same excitation signal is applied to the first audio input system and the second audio input system to generate first audio data and second audio data respectively, obtain a first quality evaluation result of the first audio data, and obtain a second quality evaluation result after noise reduction processing by performing noise reduction processing on the second audio data; based on the first audio data and the first quality evaluation result, and the second audio data and the second quality evaluation result, the noise reduction parameters used in the noise reduction processing of the second audio input system are optimized, and by migrating the noise reduction parameters that have been adjusted to the optimal effect in the first audio input system to the second audio input system, the optimization efficiency of the noise reduction parameters is greatly improved, the optimization time of the noise reduction parameters is reduced, the cost is low, and the noise reduction effect and the robustness of the system can be further improved.

Description

Technical field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular to a noise reduction parameter optimization method, apparatus, terminal device, and medium. [Background Technology]

[0002] At present, audio input systems such as microphones have been widely used in various electronic devices such as mobile phones, smart watches, Bluetooth headsets, etc.

[0003] Since input audio signals are often accompanied by noise, noise reduction is often required for the audio input of electronic devices to achieve the required audio quality and ensure recording and call quality. However, due to consistency issues in audio input systems, such as hardware consistency, it is impossible to achieve noise reduction across different audio input systems using a fixed set of parameters.

[0004] To ensure optimal sound quality for each audio input system, existing techniques involve manually calibrating noise reduction parameters within a test environment to fine-tune the system to be optimized. This noise reduction parameter generation method is not only inefficient but also requires expensive anechoic chamber construction.

[0005] Therefore, the audio noise reduction solution of the audio input system in the prior art has the problems of low efficiency and high cost. [Summary of the invention]

[0006] The embodiments of the present application provide a noise reduction parameter optimization method, apparatus, terminal device, and medium, which improve the optimization efficiency of noise reduction parameters, reduce the noise reduction parameter optimization time, are low-cost, and can further improve the noise reduction effect and enhance the robustness of the system.

[0007] In a first aspect, an embodiment of the present application provides a noise reduction parameter optimization method, comprising: applying the same excitation signal to a first audio input system and a second audio input system, so that the first audio input system and the second audio input system generate first audio data and second audio data, respectively, the first audio input system is an optimized audio input system, and the second audio input system is an audio input system to be optimized; obtaining a first quality assessment result of the first audio data, and obtaining a second quality assessment result of the second audio data after noise reduction processing after performing noise reduction processing on the second audio data; optimizing the noise reduction parameters used in the noise reduction processing of the second audio input system based on the first audio data and the first quality assessment result, as well as the second audio data and the second quality assessment result.

[0008] In the above-mentioned noise reduction parameter optimization method, the first audio input system is regarded as the original knowledge domain, and the noise reduction parameter adjustment of the original knowledge domain has reached the optimal state. Therefore, the original data is borrowed through transfer learning, and the noise reduction parameters that have been adjusted to the optimal effect are migrated to the second audio input system, realizing the process of migrating the original calibration data to the target domain, which greatly improves the optimization efficiency of the noise reduction parameters, reduces the optimization time of the noise reduction parameters, is low-cost, and can further improve the noise reduction effect and improve the robustness of the system.

[0009] In one possible implementation, obtaining a first quality assessment result of the first audio data and performing noise reduction processing on the second audio data to obtain a second quality assessment result of the second audio data after noise reduction processing include: calculating the average opinion score of the first audio data to obtain a first average opinion score; performing noise reduction processing on the second audio data to obtain tuned audio data, calculating the tuning score of the second audio data to obtain a second average opinion score; optimizing the noise reduction parameters used in the noise reduction processing of the second audio input system based on the first audio data and the first quality assessment result, as well as the second audio data and the second quality assessment result, including: constructing source domain data with the first audio data as a feature and the first average opinion score as a label; constructing target domain data with the second audio data as a feature and the second average opinion score as a label; and optimizing the noise reduction parameters of the second audio input system based on the source domain data and the target domain data using a transfer learning algorithm.

[0010] In one possible implementation, each of the noise reduction parameters is taken as a position to be optimized, and the hyperparameters of the whale optimization algorithm are set; and the whale optimization algorithm is used iteratively to optimize all noise reduction parameters simultaneously until the optimal noise reduction parameter is obtained.

[0011] In one possible implementation, the noise reduction parameters include at least one of the following: noise power spectrum estimation, noise power spectrum estimation smoothing parameter, compensation parameter when estimating noise power spectrum using minimum method, gain of Wiener filter, probability of speech absence, probability of speech presence, and noise spectrum estimation.

[0012] In a second aspect, an embodiment of the present application provides a noise reduction parameter optimization device, which is arranged in a terminal device, and the optimization device includes: an excitation module, which is used to apply the same excitation signal to a first audio input system and a second audio input system, so that the first audio input system and the second audio input system generate first audio data and second audio data, respectively, the first audio input system is an optimized audio input system, and the second audio input system is an audio input system to be optimized; an evaluation module, which is used to obtain a first quality evaluation result of the first audio data, and obtain a second quality evaluation result of the second audio data after noise reduction processing after performing noise reduction processing on the second audio data; an optimization module, which is used to optimize the noise reduction parameters used in the noise reduction processing of the second audio input system based on the first audio data and the first quality evaluation result, as well as the second audio data and the second quality evaluation result.

[0013] In one possible implementation, the evaluation module includes: a first calculation module, used to calculate the average opinion score of the first audio data to obtain a first average opinion score; a second calculation module, used to perform noise reduction processing on the second audio data to obtain tuned audio data, calculate the tuning score of the second audio data, and obtain a second average opinion score; the optimization module includes: a source domain construction module, used to construct source domain data using the first audio data as a feature and the first average opinion score as a label; a target domain construction module, used to construct target domain data using the second audio data as a feature and the second average opinion score as a label; a transfer learning module, which uses a transfer learning algorithm to optimize the noise reduction parameters of the second audio input system based on the source domain data and the target domain data.

[0014] In one possible implementation, the device further includes: a setting module for taking each of the noise reduction parameters as a position to be optimized and setting the hyperparameters of the whale optimization algorithm; an iteration module for iteratively using the whale optimization algorithm to simultaneously optimize all noise reduction parameters until the optimal noise reduction parameters are obtained.

[0015] In one possible implementation, the noise reduction parameters include at least one of the following: noise power spectrum estimation, noise power spectrum estimation smoothing parameter, compensation parameter when estimating noise power spectrum using minimum method, gain of Wiener filter, probability of speech absence, probability of speech presence, and noise spectrum estimation.

[0016] In a third aspect, an embodiment of the present application provides a terminal device comprising: at least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the method provided in the first aspect.

[0017] In a fourth aspect, an embodiment of the present application provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method provided in the first aspect.

[0018] It should be understood that the second to fourth aspects of the embodiments of the present application are consistent with the technical solutions of the first aspect of the embodiments of the present application, and the beneficial effects achieved by each aspect and the corresponding feasible implementation methods are similar and will not be repeated.

Brief Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 A flow chart of a noise reduction parameter optimization method provided in one embodiment of the present application;

[0021] Figure 2 A flowchart of a noise reduction parameter optimization method provided in another embodiment of the present application;

[0022] Figure 3 A flowchart of a noise reduction parameter optimization method provided in yet another embodiment of the present application;

[0023] Figure 4 A schematic diagram of the structure of a noise reduction parameter optimization device provided in one embodiment of this specification;

[0024] Figure 5 A schematic diagram of the structure of an evaluation module provided in one embodiment of this specification;

[0025] Figure 6 A schematic diagram of the structure of an optimization module provided in one embodiment of this specification;

[0026] Figure 7 This is a structural diagram of a noise reduction parameter optimization device provided in another embodiment of this specification. [Specific implementation method]

[0027] In order to better understand the technical solutions of this specification, the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0028] It should be clear that the embodiments described are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this specification.

[0029] The terms used in the examples of this application are for the purpose of describing specific embodiments only and are not intended to limit this specification. The singular forms "a," "an," "the," and "the" used in the examples of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0030] In existing technologies, the audio input systems in products like mobile phones, smartwatches, and Bluetooth headsets often receive noise when receiving audio signals. Therefore, noise reduction is crucial for ensuring recording and call quality. Currently, the primary approach involves setting up a test environment, manually adjusting the noise reduction parameters of the audio input system, and then testing the voice quality using equipment.

[0031] Existing methods not only require high costs to build an anechoic chamber for the test environment, but also have the problem that during the test process, the audio input system is affected by factors such as the environment and electromagnetic interference, resulting in greatly varying noise reduction index values.

[0032] Moreover, parameter adjustment usually affects the overall effect because of the adjustment of a certain parameter. Adjustment based on experience usually cannot achieve the optimal effect. Therefore, it is crucial to realize automatic parameter adjustment of audio noise reduction effect. The manual adjustment process will take a lot of time and is not conducive to quickly solving the noise interference in the hardware system.

[0033] This method cannot guarantee that many parameters are in the optimal audio effect state, and at the same time increases labor costs, is inefficient, and takes a long time.

[0034] Based on the above problems, an embodiment of the present application provides a noise reduction parameter optimization method, which uses the data of the first audio input system as existing knowledge, and migrates the noise reduction parameters that have been adjusted to the optimal effect in the existing knowledge to the second audio input system, greatly improving the optimization efficiency of the noise reduction parameters, reducing the noise reduction parameter optimization time, and having low cost. It can also further improve the noise reduction effect and enhance the robustness of the system.

[0035] Figure 1 A flow chart of a noise reduction parameter optimization method provided in one embodiment of the present application is shown in FIG. Figure 1 As shown, the above-mentioned noise reduction parameter optimization method may include:

[0036] Step 101: Apply the same excitation signal to a first audio input system and a second audio input system, so that the first audio input system and the second audio input system generate first audio data and second audio data, respectively. The first audio input system is an optimized audio input system, and the second audio input system is an audio input system to be optimized.

[0037] Two audio input systems are selected. One of these systems has already undergone noise reduction parameter optimization (referred to as the first audio input system in this embodiment), and the other is in preparation for noise reduction parameter optimization (referred to as the second audio input system in this embodiment). The two audio input systems can be different products, or different batches, specifications, models, or production runs of the same product.

[0038] An audio input system refers to a device that can convert sound into electrical signals, such as a microphone. In some cases, an audio input system can also refer to a system equipped with an audio input device, such as a mobile phone equipped with a microphone.

[0039] The same excitation signal is applied to the two selected audio input systems to generate audio data. The excitation signal can be a long tone, a short tone, interference, different tones, frequencies or a combination thereof.

[0040] Step 102: Obtain a first quality evaluation result of the first audio data, and perform noise reduction processing on the second audio data to obtain a second quality evaluation result of the noise-reduced second audio data.

[0041] A quality assessment is performed on the first audio data generated by the first audio input system to obtain a first quality assessment result.

[0042] The second audio data generated by the second audio input system is subjected to noise reduction processing. The second audio input system is a system to be prepared for noise reduction parameter optimization. That is to say, the second audio input system has not been adjusted, and the second audio data generated by it cannot meet the requirements. Therefore, it is necessary to perform noise reduction processing on it, and perform quality evaluation on the second audio data after noise reduction processing to obtain a second quality evaluation result.

[0043] Step 103: Optimize noise reduction parameters used in the noise reduction process of the second audio input system according to the first audio data and the first quality assessment result, and the second audio data and the second quality assessment result.

[0044] The first noise reduction parameter of the first audio input system can be obtained through the first audio data and the first quality assessment result. The first noise reduction parameter is used to optimize the second audio data to obtain a predicted quality assessment result. The difference between the predicted quality assessment result and the second quality assessment result is calculated. The first noise reduction parameter is adjusted according to the difference until the difference between the new predicted quality assessment result and the second quality assessment result is less than a preset threshold. The noise reduction parameter obtained at this time is the noise reduction parameter required by the second audio input system.

[0045] The above-mentioned noise reduction parameter optimization method takes the first audio input system as the original knowledge domain. The noise reduction parameter adjustment of the original knowledge domain has reached the optimal state. Therefore, the original data is borrowed through transfer learning, and the noise reduction parameters that have been adjusted to the optimal effect are migrated to the second audio input system, realizing the process of migrating the original calibration data to the target domain, greatly improving the optimization efficiency of the noise reduction parameters, reducing the noise reduction parameter optimization time, low cost, and can further improve the noise reduction effect and improve the robustness of the system.

[0046] The noise reduction parameter optimization method of the present application can effectively suppress non-steady-state noise and quickly converge and attenuate burst noise, thereby improving voice call quality.

[0047] Figure 2 A flowchart of a noise reduction parameter optimization method provided in another embodiment of the present application is shown in FIG. Figure 2 As shown, this application Figure 1 In the illustrated embodiment, step 102 further includes:

[0048] Step 201: Calculate the mean opinion score of the first audio data to obtain a first mean opinion score.

[0049] A mean opinion score (MOS) is calculated for the first audio data generated by the first audio input system. The mean opinion score (MOS) is a measurement used in the fields of quality of experience and telecommunications engineering, representing the overall quality of a stimulus or system. MOS is the arithmetic mean of all individual scores. MOS ratings are typically collected in subjective quality assessment tests, but they can also be estimated algorithmically. This application does not limit the method for calculating MOS.

[0050] Step 202: Perform noise reduction processing on the second audio data to obtain tuned audio data, calculate a tuning score of the second audio data, and obtain a second mean opinion score.

[0051] The second audio data generated by the second audio input system is subjected to noise reduction processing. The second audio input system is a system to be prepared for noise reduction parameter optimization. That is to say, the second audio input system has not been adjusted, and its average opinion score cannot meet the requirements. Therefore, it is necessary to perform noise reduction processing on it to obtain tuned audio data, and calculate the average opinion score of the tuned audio data.

[0052] The MOS score is usually required to be no less than 3.5. With the development of mobile communications, it has gradually increased to 4. Quantitative evaluation standards make parameter adjustment more digital. This application sets a MOS value greater than 4.2 as the ultimate goal of noise reduction parameter adjustment, that is, the MOS score of the second tuned audio data is greater than 4.2.

[0053] Furthermore, step 103 further includes:

[0054] Step 203 : construct source domain data using the first audio data as features and the first mean opinion score as a label.

[0055] The first audio data is used as a feature and the first mean opinion score is used as a label to construct source domain data.

[0056] Step 204 : construct target domain data using the second audio data as features and the second mean opinion score as a label.

[0057] The second audio data is used as features and the second mean opinion score is used as a label to construct target domain data.

[0058] Step 205: Based on the source domain and the target domain, a transfer learning algorithm is used to optimize the noise reduction parameters of the second audio input system.

[0059] For example, construct the training sample set and test sample set of the source domain and the target domain. Assume that the training dataset and the test dataset are T and S respectively, as shown below:

[0060]

[0061] Among them, T t and T s are the training sample sets of the target domain and the source domain respectively, x i is the feature vector of the i-th sample, y i is the corresponding parameter label, n and m are the number of samples in the training sample set of the target domain and the source domain respectively, generally n≤m. The test set is selected from different data from the training set.

[0062] Input training samples and train the learning model. The objective function is obtained through training with the training set, and then used to predict the label of new data. The transfer learning decision function expression is:

[0063]

[0064] Input test samples, tune the parameter data, and verify the performance of the model.

[0065] Preferably, the noise reduction parameter includes at least one of the following:

[0066] Noise power spectrum estimation, smoothing parameter for noise power spectrum estimation, compensation parameter when estimating noise power spectrum using the minimum method, gain of the Wiener filter, probability of speech absence, probability of speech presence, noise spectrum estimation.

[0067] Many parameters in the audio input system interact with each other. Adjusting a single parameter cannot achieve the optimal effect. Only by adjusting all parameters at the same time can the noise reduction effect be optimized.

[0068] In this application, the noise reduction system uses the frequency domain segmented signal-to-noise ratio indicator for evaluation, and optimizes each parameter value through parameter adjustment. This is achieved by adjusting the full-band parameters in the speech frame parameters and noise frame parameters. The optimization parameters are as follows: noise power spectrum estimation, noise power spectrum estimation smoothing parameter, compensation parameter when estimating the noise power spectrum using the minimum value method, Wiener filter gain, speech absence probability, speech presence probability, and noise spectrum estimation.

[0069] Minimum noise power spectrum estimation

[0070]

[0071] Wiener filter gain

[0072]

[0073]

[0074]

[0075] Compensation parameters when estimating noise power spectrum using the minimum method

[0076]

[0077]

[0078] Probability of speech absence

[0079]

[0080] Calculate the probability of speech presence

[0081]

[0082]

[0083] Get the noise spectrum estimate

[0084]

[0085] α D (k,λ)=α d +(1-α d )p(k,λ)

[0086] In the above formula, λ represents the number of frames, κ represents the frequency index, Υ represents the noisy speech spectrum in the frequency domain, σ represents the noise spectrum, and α represents the smoothing factor. The noise spectrum estimation algorithm seeks to determine the time-frequency-dependent smoothing factor, which can then be used to estimate the noise using the above formula. The smoothing factor can be calculated based on the signal-to-noise ratio or a fixed value. However, it is more commonly calculated based on the probability of speech presence or absence at frequency k. The relationship between this smoothing factor and the probability of speech presence can be shown later.

[0087] Figure 3 A flowchart of a noise reduction parameter optimization method provided in another embodiment of the present application is shown in FIG. Figure 3 As shown, this application Figure 1 In the illustrated embodiment, the method further comprises:

[0088] In step 301, each of the noise reduction parameters is taken as a position to be optimized, and the hyperparameters of the whale optimization algorithm are set.

[0089] Set each parameter in the denoising process to a position X(t) in the whale optimization algorithm, and set the hyperparameters of the whale optimization algorithm, such as the size of the whale population and the maximum number of iterations.

[0090] Step 302, iteratively, using the whale optimization algorithm to optimize all noise reduction parameters simultaneously until the optimal noise reduction parameters are obtained.

[0091] Initialize the position of the whale population, calculate the corresponding fitness value of each whale, sort them according to the size of the fitness value, and select n as the initial population. Perform iterations: calculate the size of the n individual fitness values, find the position of the individual with the smallest fitness value as the optimal position, and update the position of the next generation; if the termination condition is met, output the optimal individual, that is, find the optimal solution for the parameters.

[0092] By iteratively optimizing each parameter, the difference between the optimal value and the parameter setting is calculated to continuously approach the optimal value. All parameters are adjusted to the optimal value.

[0093] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0094] Figure 4 This is a schematic diagram of the structure of a noise reduction parameter optimization device provided by an embodiment of the present invention. The noise reduction parameter optimization device is provided in a terminal device, such as Figure 4 As shown, the noise reduction parameter optimization device may include: an excitation module 41, an evaluation module 42 and an optimization module 43;

[0095] The excitation module 41 is configured to apply the same excitation signal to the first audio input system and the second audio input system, so that the first audio input system and the second audio input system generate first audio data and second audio data, respectively. The first audio input system is an optimized audio input system, and the second audio input system is an audio input system to be optimized.

[0096] an evaluation module 42 configured to obtain a first quality evaluation result of the first audio data, and to obtain a second quality evaluation result of the noise-reduced second audio data after performing noise reduction processing on the second audio data;

[0097] The optimization module 43 is configured to optimize noise reduction parameters used in the noise reduction process of the second audio input system according to the first audio data and the first quality evaluation result, and the second audio data and the second quality evaluation result.

[0098] Figure 4 The noise reduction parameter optimization device provided in the embodiment shown can be used to implement the present invention. Figure 1 The technical solution of the method embodiment shown, its implementation principle and technical effects can be further referred to the relevant description in the method embodiment.

[0099] In one embodiment of the present application, Figure 5 A schematic diagram of the structure of an evaluation module provided by an embodiment of the present invention is shown in FIG. Figure 5 As shown, the evaluation module 42 includes: a first calculation module 421 and a second calculation module 422;

[0100] The first calculation module 421 is configured to calculate a mean opinion score of the first audio data to obtain a first mean opinion score.

[0101] A second calculation module 422 is configured to perform noise reduction processing on the second audio data to obtain tuned audio data, calculate a tuning score for the second audio data, and obtain a second mean opinion score;

[0102] Further, Figure 6 A schematic diagram of the structure of an optimization module provided by an embodiment of the present invention is shown in FIG. Figure 6 As shown, the above optimization module 43 includes: a source domain construction module 431, a target domain construction module 432 and a transfer learning module 433;

[0103] The source domain construction module 431 is configured to construct source domain data using the first audio data as a feature and the first mean opinion score as a label;

[0104] A target domain construction module 432 is configured to construct target domain data using the second audio data as features and the second mean opinion score as a label;

[0105] The transfer learning module 433 optimizes the noise reduction parameters of the second audio input system using a transfer learning algorithm based on the source domain data and the target domain data.

[0106] Furthermore, the noise reduction parameter optimization device may further include: a setting module 71 and an iteration module 72;

[0107] A setting module 71 is used to set the hyperparameters of the whale optimization algorithm by taking each of the noise reduction parameters as a position to be optimized;

[0108] The iterative module 72 is used for iteratively optimizing all noise reduction parameters simultaneously using the whale optimization algorithm until the optimal noise reduction parameters are obtained.

[0109] The noise reduction parameter optimization device provided in the above embodiment can be used to implement the present application Figures 1 to 3 The technical solution of the method embodiment shown, its implementation principle and technical effects can be further referred to the relevant description in the method embodiment.

[0110] The embodiment of the present application provides a terminal device, which may include at least one processor; and at least one memory in communication with the processor, wherein: the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the instructions of this specification. Figures 1 to 3 The embodiment shown provides a method for optimizing noise reduction parameters.

[0111] The terminal device may be a smart electronic device such as a smartphone, a tablet computer, or a laptop computer. This embodiment does not limit the form of the terminal device.

[0112] The embodiment of the present application provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute the instructions of the present invention. Figures 1 to 3 The embodiment shown provides a method for optimizing noise reduction parameters.

[0113] The above-mentioned non-transitory computer-readable storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM) or a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by an instruction execution system, device or device or used in combination with it.

[0114] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0115] Program code embodied on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, radio frequency (RF), etc., or any suitable combination of the foregoing.

[0116] Computer program code for performing the operations of this specification may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0117] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0118] In the description of the embodiments of the present invention, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of this specification. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0119] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout this specification, "plurality" means at least two, such as two or three, unless otherwise specifically defined.

[0120] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of this specification includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of this specification belong.

[0121] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0122] It should be noted that the terminals involved in the embodiments of the present application may include but are not limited to personal computers (PCs), personal digital assistants (PDAs), wireless handheld devices, tablet computers, mobile phones, MP3 players, MP4 players, etc.

[0123] In the several embodiments provided in this specification, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0124] In addition, the functional units in the various embodiments of this specification may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional units.

[0125] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform some steps of the method described in various embodiments of this specification. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0126] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.

Claims

1. A noise reduction parameter optimization method, characterized in that: include: Applying the same excitation signal to a first audio input system and a second audio input system, so that the first audio input system and the second audio input system generate first audio data and second audio data, respectively, wherein the first audio input system is an optimized audio input system and the second audio input system is an audio input system to be optimized; Obtaining a first quality assessment result of the first audio data, and performing noise reduction processing on the second audio data to obtain a second quality assessment result of the noise-reduced second audio data; Optimizing noise reduction parameters used in noise reduction processing performed by the second audio input system based on the first audio data and the first quality assessment result, and the second audio data and the second quality assessment result; After obtaining the first quality assessment result and the second quality assessment result, the method further includes: obtaining noise reduction parameters of the second audio input system according to the first quality assessment result and the second quality assessment result, including: determining a first noise reduction parameter of a first audio input system according to the first audio data and the first quality assessment result; Optimizing the second audio data using the first noise reduction parameter to obtain a predicted quality assessment result; The first noise reduction parameter is adjusted based on the difference between the predicted quality assessment result and the second quality assessment result, and the adjustment is completed when the difference is less than a preset threshold, and the first noise reduction parameter at this time is determined as the noise reduction parameter.

2. The method according to claim 1, wherein: Obtaining a first quality assessment result of the first audio data, and performing noise reduction processing on the second audio data to obtain a second quality assessment result of the noise-reduced second audio data, comprising: Calculating a mean opinion score of the first audio data to obtain a first mean opinion score, and using the first mean opinion score as a first quality assessment result of the first audio data; Performing noise reduction processing on the second audio data to obtain tuned audio data, calculating a tuning score for the second audio data to obtain a second mean opinion score, and using the second mean opinion score as a second quality assessment result of the second audio data; Optimizing noise reduction parameters used in noise reduction processing performed on the second audio input system based on the first audio data and the first quality assessment result, and the second audio data and the second quality assessment result, includes: constructing source domain data using the first audio data as features and the first mean opinion score as a label; constructing target domain data using the second audio data as features and the second mean opinion score as a label; Based on the source domain data and the target domain data, a transfer learning algorithm is used to optimize the noise reduction parameters of the second audio input system.

3. The method according to claim 1, characterized in that The method further comprises: Taking each of the noise reduction parameters as a position to be optimized, setting the hyperparameters of the whale optimization algorithm; Iteratively, all noise reduction parameters are optimized simultaneously using the whale optimization algorithm until the optimal noise reduction parameters are obtained.

4. The method according to claim 1, wherein The noise reduction parameters include at least one of the following: Noise power spectrum estimation, smoothing parameter for noise power spectrum estimation, compensation parameter when estimating noise power spectrum using the minimum method, gain of the Wiener filter, probability of speech absence, probability of speech presence, noise spectrum estimation.

5. A noise reduction parameter optimization device, provided in a terminal device, characterized in that: The optimization device comprises: an excitation module, configured to apply the same excitation signal to a first audio input system and a second audio input system, so that the first audio input system and the second audio input system generate first audio data and second audio data, respectively; the first audio input system is an optimized audio input system, and the second audio input system is an audio input system to be optimized; an evaluation module, configured to obtain a first quality evaluation result of the first audio data, and to obtain a second quality evaluation result of the noise-reduced second audio data after performing noise reduction processing on the second audio data; an optimization module, configured to optimize noise reduction parameters used in a noise reduction process performed by the second audio input system based on the first audio data and the first quality assessment result, and the second audio data and the second quality assessment result; After obtaining the first quality assessment result and the second quality assessment result, the method further includes: obtaining noise reduction parameters of the second audio input system according to the first quality assessment result and the second quality assessment result, including: determining a first noise reduction parameter of a first audio input system according to the first audio data and the first quality assessment result; Optimizing the second audio data using the first noise reduction parameter to obtain a predicted quality assessment result; The first noise reduction parameter is adjusted based on the difference between the predicted quality assessment result and the second quality assessment result, and the adjustment is completed when the difference is less than a preset threshold, and the first noise reduction parameter at this time is determined as the noise reduction parameter.

6. The device according to claim 5, characterized in that: The evaluation module includes: a first calculation module, configured to calculate a mean opinion score of the first audio data to obtain a first mean opinion score, and use the first mean opinion score as a first quality assessment result of the first audio data; a second computing module, configured to perform noise reduction processing on the second audio data to obtain tuned audio data, calculate a tuning score for the second audio data to obtain a second mean opinion score, and use the second mean opinion score as a second quality assessment result of the second audio data; The optimization module includes: A source domain construction module, configured to construct source domain data using the first audio data as a feature and the first mean opinion score as a label; a target domain construction module, configured to construct target domain data using the second audio data as a feature and the second mean opinion score as a label; A transfer learning module optimizes the noise reduction parameters of the second audio input system using a transfer learning algorithm based on the source domain data and the target domain data.

7. The device according to claim 5, characterized in that The device further comprises: A setting module, used to treat each of the noise reduction parameters as a position to be optimized and set the hyperparameters of the whale optimization algorithm; The iterative module is used for iteratively optimizing all noise reduction parameters simultaneously using the whale optimization algorithm until the optimal noise reduction parameters are obtained.

8. The device according to claim 5, characterized in that The noise reduction parameters include at least one of the following: Noise power spectrum estimation, smoothing parameter for noise power spectrum estimation, compensation parameter when estimating noise power spectrum using the minimum method, gain of the Wiener filter, probability of speech absence, probability of speech presence, noise spectrum estimation.

9. A terminal device comprising: at least one processor; as well as at least one memory in communication with the processor, wherein: The memory stores program instructions that can be executed by the processor, and the processor can execute the method according to any one of claims 1 to 4 by calling the program instructions. 10 . A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions enable the computer to execute the method according to claim 1 .

Images