A noise elimination test method, system, audio and video equipment and storage medium

Abstract

The present application relates to a noise elimination test method, system, audio and video equipment and storage medium, which obtains initial audio information, and preliminarily processes the initial audio information to obtain test audio information; divides the speech sequence of the test audio information into multiple test speech frames, and performs noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information; tests the intermediate audio information through an audio test model, and evaluates the noise elimination effect of the test audio according to the test results. The present invention can accurately test the noise elimination effect of VOIP equipment during operation, solves the problem of low efficiency and reliability when the original method that relies on subjective judgment verifies and tests the noise elimination effect of each relevant scene one by one, and meets the actual application needs.

Description

Technical Field

[0001] The present application relates to the technical field of speech signal processing, and in particular to a noise elimination test method, system, audio and video equipment, and storage medium. Background Art

[0002] Communication is the exchange and transmission of information between people through some kind of media. A network is a data link that connects isolated workstations or hosts together with physical links. A communication network is a link that physically connects isolated devices to achieve information exchange between people, people and computers, and computers and computers, thereby achieving the purpose of resource sharing and communication.

[0003] As an important part of network communication, audio and video conferencing plays a vital role in conference communication. An audio and video conferencing system refers to a device that enables two or more individuals or groups in different places to transmit sound, images and file data to each other through transmission lines and multimedia equipment to achieve instant and interactive communication and to achieve simultaneous conferences, such as VOIP equipment.

[0004] As we all know, there is usually noise when VOIP equipment is working, so it needs to be eliminated to improve the quality of audio and video conferences. However, the noise elimination results of existing VOIP equipment are usually judged by subjective judgment methods. As the number of VOIP equipment models continues to increase, the number of acquisition and playback devices and scenarios they carry also increases. When a problem is modified, the original method that relies on subjective judgment needs to verify and test each related scenario one by one, which is inefficient and unreliable. Summary of the invention

[0005] Based on this, it is necessary to provide a noise elimination test method, system, audio and video equipment and storage medium to address the above technical problems. This method can accurately test the noise elimination effect of the VOIP device during operation, and solves the problem of low efficiency and reliability when the original method relies on subjective judgment to verify the noise elimination effect of each relevant scene one by one.

[0006] In a first aspect, an embodiment of the present invention provides a noise elimination test method, wherein the method comprises the following steps:

[0007] Acquiring initial audio information, and performing preliminary processing on the initial audio information to obtain test audio information;

[0008] Dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information;

[0009] The intermediate audio information is tested by an audio test model, and the noise elimination effect of the test audio is evaluated according to the test result.

[0010] Furthermore, the method of preliminarily processing the initial audio information to obtain the test audio information includes:

[0011] Determining the type of the current audio signal according to the initial audio information;

[0012] When the type of the current audio signal is scene audio, testing the scene audio according to the target audio information, and using the scene audio that passes the test as the test audio;

[0013] When the type of the current audio signal is pure audio, reverberation processing is performed on the pure audio and the reference audio, and the delayed reverberation reference audio is mixed with the reverberation pure audio to obtain the test audio.

[0014] Furthermore, the method for testing the intermediate audio information by using an audio test model includes:

[0015] Dividing the intermediate audio of the intermediate audio information into a plurality of intermediate speech frames;

[0016] Calculating the speech energy of the intermediate speech frame by using an energy model, and smoothing the speech energy of the intermediate speech frame and the energy of the historical frame to obtain the energy value of the intermediate audio information;

[0017] The energy model is:

[0018]

[0019] RMS i =αRMS cur +(1-α)RMS i-1

[0020] In the formula, the speech data of the i-th microphone is represented by x i1 , x i2 , …, x iL ; L is the speech frame length.

[0021] Furthermore, the method further comprises:

[0022] The intermediate audio and the corresponding clean audio are respectively divided into a plurality of intermediate speech frames and clean speech frames, and the silence frames therein are deleted to obtain intermediate data frames and clean data frames;

[0023] Adding a Hanning window to the intermediate data frame and the clean data frame respectively and calculating the corresponding short-time DFT signal and 1 / 3 octave time-frequency decomposition signal;

[0024] The 1 / 3 octave time-frequency decomposition signals corresponding to the intermediate data frames and the pure data frames are normalized, and the STOI value of the intermediate audio information is determined according to the obtained normalization factor, signal distortion ratio and relationship number.

[0025] Furthermore, the method further comprises:

[0026] Performing noise testing and reverberation testing on the intermediate audio information simultaneously, separately or successively;

[0027] The noise elimination effect of the test audio is evaluated according to the noise test results and the reverberation test results.

[0028] Furthermore, the model for performing noise testing on the intermediate audio information is:

[0029]

[0030] In the formula, x(n) is the pure audio, is the intermediate audio, N is the frame length, and M is the number of frames.

[0031] Furthermore, the method for performing a reverberation test on the intermediate audio includes:

[0032] After filtering the intermediate speech audio input with a 23-channel filter having an equivalent rectangular bandwidth, the instantaneous envelope of each frequency band output signal is obtained by a Hilbert transform method;

[0033] The instantaneous envelope is windowed and subjected to DFT transformation to obtain a reverberation signal;

[0034] After the reverberation signal is input into an 8-channel auditory modulation filter for processing, the speech reverberation modulation ratio of the intermediate audio is obtained according to the ratio of the energy of the first 4 frequency bands to the energy of the last 4 frequency bands, and the result of reverberation of the intermediate audio is evaluated according to the speech reverberation modulation ratio.

[0035] On the other hand, an embodiment of the present invention further provides a noise elimination test system, the system comprising:

[0036] An acquisition module, used for acquiring initial audio information and performing preliminary processing on the initial audio information to obtain test audio information;

[0037] A testing module, used for dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information;

[0038] An evaluation module is used to test the intermediate audio information through an audio test model, and evaluate the noise elimination effect of the test audio according to the test result.

[0039] An embodiment of the present invention further provides an audio and video device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0040] Acquiring initial audio information, and performing preliminary processing on the initial audio information to obtain test audio information;

[0041] Dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information;

[0042] The intermediate audio information is tested by an audio test model, and the noise elimination effect of the test audio is evaluated according to the test result.

[0043] The embodiment of the present invention further provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0044] Acquiring initial audio information, and performing preliminary processing on the initial audio information to obtain test audio information;

[0045] Dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information;

[0046] The intermediate audio information is tested by an audio test model, and the noise elimination effect of the test audio is evaluated according to the test result.

[0047] The above-mentioned noise elimination test method, system, audio and video equipment and storage medium can obtain initial audio information, and preliminarily process the initial audio information to obtain test audio information; divide the speech sequence of the test audio information into multiple test speech frames, and perform noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information; test the intermediate audio information through an audio test model, and evaluate the noise elimination effect of the test audio according to the test results, and can automatically judge the difference between the test case output results and the standard results of each scene, reduce the time for manual judgment of the results, and improve the efficiency of version release. Because the present invention accurately tests the noise elimination effect of the VOIP device during operation, it solves the problem of low efficiency and reliability when the original method that relies on subjective judgment verifies the noise elimination effect of each relevant scene one by one, and meets the actual application needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of a flow chart of a noise elimination test method provided by an embodiment of the present invention;

[0049] Figure 2 for Figure 1 Schematic diagram of the process of 101;

[0050] Figure 3 for Figure 1 Schematic diagram of the process of 103;

[0051] Figure 4 for Figure 1 A schematic flow chart of another embodiment of 103;

[0052] Figure 5 for Figure 1 A schematic flow chart of another embodiment of 103;

[0053] Figure 6 A structural block diagram of a noise elimination test system in one embodiment;

[0054] Figure 7 for Figure 6 Get the structural diagram of the module;

[0055] Figure 8 for Figure 6 The structural diagram of the evaluation module;

[0056] Fig. 9 It is a structural block diagram of an audio and video device in another embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] See also Figures 1 to 5 The noise elimination test method provided in the present application is executed by software or hardware of an audio and video device (such as a smart terminal or a server) for controlling the noise test, wherein the method comprises steps 101 to 103:

[0059] Step 101: Acquire initial audio information, and preliminarily process the initial audio information to obtain test audio information.

[0060] As mentioned above, when making a video call through a VOIP device, there are usually various noises and reverberations, so they need to be suppressed and eliminated to improve the quality of the video call. However, due to the uncertainty of the initial audio information, it is necessary to preliminarily process the initial audio information to obtain the test audio information, which provides the necessary conditions for the audio noise elimination test. At the same time, the reliability of the noise elimination model can be verified.

[0061] Furthermore, the method of preliminarily processing the initial audio information to obtain the test audio information includes:

[0062] Step 201: determine the type of the current audio signal according to the initial audio information.

[0063] As mentioned above, since the initial audio information includes audio with background sound and pure audio. There are also multiple invalid, similar or different audios in the audio with background sound, so it is necessary to perform preliminary processing on it. In addition, in order to obtain the noise elimination effect of the noise elimination model on the audio that meets the specific environment and artistic conception, it is necessary to perform noise synthesis and reverberation processing on the pure sound to obtain the corresponding test audio. It can be understood that by performing preliminary processing on the initial audio information, the test audio information that meets specific requirements is obtained, which provides the necessary conditions for testing the reliability of the noise elimination model.

[0064] Step 202: When the type of the current audio signal is scene audio, the scene audio is tested according to target audio information, and the scene audio that passes the test is used as test audio.

[0065] As described above, when the type of the current audio signal is scene audio with background sound, it is necessary to test according to the target audio corresponding to the current scene audio, and use the set of scene audios that meet the requirements as test audio, thereby preparing the test data set and providing the necessary conditions for subsequent robustness and reliability testing of the noise elimination model.

[0066] Specifically, the scene audio may include the following scenes:

[0067] Scenario 1: After the local end establishes a call, the local end speaks normally with the remote end to test the convergence speed of background noise suppression and the effect after convergence.

[0068] Scenario 2: The near end keeps talking after establishing a call, and the effect of noise cancellation is tested when it has not converged.

[0069] Scenario 3: Speak in an environment with relatively high but stable near-end background noise (air conditioner at maximum speed) to test the noise cancellation effect during speech (whether the background noise is too loud or the background noise is low but the voice fluctuates).

[0070] Scenario 4: Speak in a near-end burst noise environment (knocking on the table, mouse, keyboard, moving chairs, opening and closing doors, folding paper, etc.) to test the noise cancellation effect during speaking (whether the background noise is too loud or the background noise is small but the sound fluctuates).

[0071] Scenario 5: There is background voice, phone ringing, or music playing at the near end. Test the noise cancellation effect during this period (whether the background noise is too loud or the background noise is small but the sound fluctuates).

[0072] Scenario 6: The near end establishes a call in a conference room with high reverberation and tests the effect of reverberation reduction when different suppression parameters are set.

[0073] Step 203: When the type of the current audio signal is clean audio, reverberation processing is performed on the clean audio and the reference audio, and the delayed reverberation reference audio is mixed with the reverberation clean audio to obtain a test audio.

[0074] As described above, when the type of the current audio signal is pure audio, it is necessary to add specific noise to the pure audio according to the test requirements, that is, the pure audio and the reference audio are reverberated according to the requirements, and the reference audio is delayed by 30ms to 200ms and then mixed with the reverberated pure audio and the reference audio. A non-uniform waveform shrinkage mixing method can be used to obtain the test audio, thereby realizing the preparation of the test data set and providing the necessary conditions for the subsequent robustness and reliability testing of the noise elimination model.

[0075] Step 102: Divide the speech sequence of the test audio information into a plurality of test speech frames, and perform noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information.

[0076] Specifically, the speech sequence of the test audio information is divided into a plurality of test speech frames (such as 10ms or 8ms), and then the test speech frames are subjected to noise suppression such as ANS noise suppression method and reverberation suppression processing to obtain intermediate audio information. It can be understood that the intermediate audio information is the audio information obtained after noise suppression and reverberation suppression are performed on the test audio information. The acquisition of the intermediate audio information provides necessary conditions for the subsequent test of the audio noise elimination effect.

[0077] Step 103: Test the intermediate audio information using an audio test model, and evaluate the noise elimination effect of the test audio according to the test result.

[0078] As described above, by testing the intermediate audio information, the reliability of the noise elimination model can be evaluated according to the test result, so as to provide a basis for the subsequent adjustment and correction of the noise elimination model.

[0079] Furthermore, the method for testing the intermediate audio information by using an audio test model comprises the following steps:

[0080] Step 301: divide the intermediate audio of the intermediate audio information into a plurality of intermediate speech frames.

[0081] Step 302: Calculate the speech energy of the intermediate speech frame through an energy model, and smooth the speech energy of the intermediate speech frame and the energy of the historical frame to obtain the energy value of the intermediate audio information.

[0082] The energy model is:

[0083]

[0084] RMS i =αRMS cur +(1-α)RMS i-1

[0085] In the formula, the speech data of the i-th microphone is represented by x i1 , x i2 , …, x iL ; L is the speech frame length and α is the smoothing coefficient, which is usually taken as 0.1.

[0086] As described above, since the intermediate audio in the intermediate audio information is relatively long and there is uncertainty and imbalance in the noise, it is necessary to divide it into several speech frames, and calculate its speech energy through the energy model, and finally smooth the speech energy of the intermediate speech frame and the energy of the historical frame to obtain the energy value of the intermediate audio information. If the difference between the obtained energy value and the standard energy value is less than 3dB, the robustness and noise elimination effect of the noise elimination model are not up to standard. Among them, the difference range between the energy value and the standard energy value can also be adjusted according to actual needs. It can be understood that in other embodiments, the noise elimination effect of the test audio can also be evaluated by an objective speech quality evaluation method, that is, according to the objective score of the intermediate signal.

[0087] Furthermore, the method for testing the intermediate audio information by using an audio test model comprises the following steps:

[0088] Step 401 : divide the intermediate audio and the corresponding clean audio into a plurality of intermediate speech frames and clean speech frames respectively, and delete the silent frames therein to obtain intermediate data frames and clean data frames.

[0089] Step 402 , adding a Hanning window to the intermediate data frame and the clean data frame respectively and calculating to obtain the corresponding short-time DFT signal and 1 / 3 octave time-frequency decomposition signal.

[0090] Step 403, normalizing the 1 / 3 octave time-frequency decomposition signals corresponding to the intermediate data frame and the clean data frame, and determining the STOI value of the intermediate audio information according to the obtained normalization factor, signal distortion ratio and relationship coefficient. The signal distortion ratio is used to express the distortion rate of the intermediate data frame, and the relationship coefficient is used to express the correlation between the intermediate data frame and the signal distortion ratio.

[0091] As described above, the reliability of the noise test is improved by dividing the intermediate audio and the corresponding clean audio into multiple intermediate speech frames and clean speech frames; the efficiency of the noise cancellation test is improved by deleting the silent frames in the intermediate speech frames and clean speech frames to reduce invalid audio tests; the 1 / 3 octave time-frequency decomposition signals corresponding to the intermediate data frames and the clean data frames are normalized to facilitate calculation of the signal distortion ratio of the intermediate data frames based on the obtained normalization factors and the 1 / 3 octave time-frequency decomposition signals of the intermediate data frames and the clean data frames; and the STOI value of the intermediate audio information is calculated to facilitate evaluation of the noise elimination effect of the test audio based on the STOI value.

[0092] Furthermore, the method for testing the intermediate audio information by using an audio test model comprises the following steps:

[0093] Step 501: Perform a noise test and a reverberation test on the intermediate audio information simultaneously, separately or successively.

[0094] Specifically, the model for performing noise testing on the intermediate audio information is:

[0095]

[0096] In the formula, x(n) is the pure audio, is the intermediate audio, N is the frame length, and M is the number of frames. When the mean value needs to be calculated, only frames with SNR in the range of [-10,35]dB are considered.

[0097] Specifically, the method for performing a reverberation test on the intermediate audio includes:

[0098] After filtering the intermediate speech audio input with a 23-channel filter having an equivalent rectangular bandwidth, the instantaneous envelope of each frequency band output signal is obtained by a Hilbert transform method;

[0099] The instantaneous envelope is windowed and subjected to DFT transformation to obtain a reverberation signal;

[0100] After the reverberation signal is input into an 8-channel auditory modulation filter for processing, the speech reverberation modulation ratio of the intermediate audio is obtained according to the ratio of the energy of the first 4 frequency bands to the energy of the last 4 frequency bands, and the result of reverberation of the intermediate audio is evaluated according to the speech reverberation modulation ratio.

[0101] Step 502: The noise elimination effect of the test audio is evaluated according to the noise test result and the reverberation test result.

[0102] Specifically, the noise elimination effect of the test audio can be evaluated according to the difference between the noise test result and the reverberation test result and the corresponding standard result, and can also be evaluated according to the correlation between the noise test result and the reverberation test result.

[0103] The above-mentioned noise elimination test method can obtain initial audio information, and preliminarily process the initial audio information to obtain test audio information; divide the speech sequence of the test audio information into multiple test speech frames, and perform noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information; test the intermediate audio information through an audio test model, and evaluate the noise elimination effect of the test audio according to the test results, and can automatically determine the difference between the test case output results and the standard results of each scene, reduce the time for manual judgment of the results, and improve the efficiency of version release. Since the present invention accurately tests the noise elimination effect of the VOIP device during operation, it solves the problem of low efficiency and reliability when the original method that relies on subjective judgment verifies the noise elimination effect of each relevant scene one by one, and meets the actual application needs.

[0104] It should be understood that, although the various steps in the above-mentioned flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above-mentioned flow chart may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0105] In one embodiment, Figure 6 As shown, a noise elimination test system is provided, and the system includes: an acquisition module 10, a test module 20, and an evaluation module 30.

[0106] The acquisition module 10 is used to acquire initial audio information and perform preliminary processing on the initial audio information to obtain test audio information.

[0107] For further information, see Figure 7 , the acquisition module 10 includes:

[0108] The judging unit 11 is configured to judge the type of the current audio signal according to the initial audio information.

[0109] The testing unit 12 is used for testing the scene audio according to target audio information when the type of the current audio signal is scene audio, and taking the scene audio that passes the test as the test audio.

[0110] The mixing unit 13 is used for performing reverberation processing on the clean audio and the reference audio when the type of the current audio signal is clean audio, and mixing the delayed reverberation reference audio with the reverberation clean audio to obtain the test audio.

[0111] The testing module 20 is used to divide the speech sequence of the test audio information into a plurality of test speech frames, and perform noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information.

[0112] The evaluation module 30 is used to test the intermediate audio information through an audio test model, and evaluate the noise elimination effect of the test audio according to the test result.

[0113] For further information, see Figure 8 , the evaluation module 30 comprises:

[0114] First evaluation unit 31: for

[0115] Dividing the intermediate audio of the intermediate audio information into a plurality of intermediate speech frames;

[0116] Calculating the speech energy of the intermediate speech frame by using an energy model, and smoothing the speech energy of the intermediate speech frame and the energy of the historical frame to obtain the energy value of the intermediate audio information;

[0117] The energy model is:

[0118]

[0119] RMS i =αRMS cur +(1-α)RMS i-1

[0120] In the formula, the speech data of the i-th microphone is represented by x i1 , x i2 , …, x iL ; L is the speech frame length.

[0121] The second evaluation unit 32 is used for

[0122] The intermediate audio and the corresponding clean audio are respectively divided into a plurality of intermediate speech frames and clean speech frames, and the silence frames therein are deleted to obtain intermediate data frames and clean data frames;

[0123] Adding a Hanning window to the intermediate data frame and the clean data frame respectively and calculating the corresponding short-time DFT signal and 1 / 3 octave time-frequency decomposition signal;

[0124] The 1 / 3 octave time-frequency decomposition signals corresponding to the intermediate data frames and the pure data frames are normalized, and the STOI value of the intermediate audio information is determined according to the obtained normalization factor, signal distortion ratio and relationship number.

[0125] The third evaluation unit 33: used for

[0126] Performing noise testing and reverberation testing on the intermediate audio information simultaneously, separately or successively;

[0127] The model for performing noise testing on the intermediate audio information is:

[0128]

[0129] In the formula, x(n) is the pure audio, is the intermediate audio, N is the frame length, M is the number of frames, SNR∈[-11,36]dB;

[0130] The method for performing a reverberation test on the intermediate audio comprises: filtering the intermediate speech audio input into a 23-channel filter with an equivalent rectangular bandwidth, obtaining the instantaneous envelope of each frequency band output signal by a Hilbert transform method, windowing the instantaneous envelope and performing DFT transform processing to obtain a reverberation signal, inputting the reverberation signal into an 8-channel auditory modulation filter for processing, obtaining the speech reverberation modulation ratio of the intermediate audio according to the ratio of the energy of the first 4 frequency bands to the energy of the last 4 frequency bands, and evaluating the result of reverberation of the intermediate audio according to the speech reverberation modulation ratio;

[0131] The noise elimination effect of the test audio is evaluated according to the noise test results and the reverberation test results.

[0132] The above-mentioned noise elimination test system can obtain initial audio information, and preliminarily process the initial audio information to obtain test audio information; divide the speech sequence of the test audio information into multiple test speech frames, and perform noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information; test the intermediate audio information through an audio test model, and evaluate the noise elimination effect of the test audio according to the test results, and can automatically determine the difference between the test case output results and the standard results of each scene, reduce the time for manual judgment of the results, and improve the efficiency of version release. Since the present invention accurately tests the noise elimination effect of the VOIP device during operation, it solves the problem of low efficiency and reliability when the original method that relies on subjective judgment verifies the noise elimination effect of each relevant scene one by one, and meets the actual application needs.

[0133] In one embodiment, an audio and video device is provided. The audio and video device may be a server, and its internal structure diagram may be as follows: Fig. 9 As shown. The audio and video device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the audio and video device is used to provide computing and control capabilities. The memory of the audio and video device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the audio and video device is used to store data. The network interface of the audio and video device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for terminal application interaction is implemented.

[0134] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the audio and video equipment to which the scheme of the present application is applied. The specific audio and video equipment may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0135] In one embodiment, an audio and video device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0136] Acquiring initial audio information, and performing preliminary processing on the initial audio information to obtain test audio information;

[0137] Dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information;

[0138] The intermediate audio information is tested by an audio test model, and the noise elimination effect of the test audio is evaluated according to the test result.

[0139] In one embodiment, a storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0140] Acquiring initial audio information, and performing preliminary processing on the initial audio information to obtain test audio information;

[0141] Dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information;

[0142] The intermediate audio information is tested by an audio test model, and the noise elimination effect of the test audio is evaluated according to the test result.

[0143] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other storage medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0144] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0145] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means 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 the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0146] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0147] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A noise elimination test method, characterized in that: The method comprises the following steps: Acquiring initial audio information, and performing preliminary processing on the initial audio information to obtain test audio information; Dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information; Testing the intermediate audio information by using an audio test model, and evaluating the noise elimination effect of the test audio according to the test result; The method of performing preliminary processing on the initial audio information to obtain the test audio information includes: Determining the type of the current audio signal according to the initial audio information; When the type of the current audio signal is scene audio, testing the scene audio according to the target audio information, and using the scene audio that passes the test as the first test audio; When the type of the current audio signal is pure audio, reverberation processing is performed on the pure audio and the reference audio, and the delayed reverberation reference audio is mixed with the reverberation pure audio to obtain a second test audio; The testing the scene audio according to the target audio information includes: After the near end establishes a call, speak alternately with the far end normally to test the convergence speed of background noise suppression and the effect after convergence; After the near end establishes a call, keep talking to test the effect of noise cancellation when it has not converged; Speak in an environment with relatively loud but stable near-end background noise to test the noise cancellation effect during speaking; Speak in a near-end burst noise environment to test the noise cancellation effect during speaking; There is background voice, phone ringing, or music playing at the near end. Test the noise cancellation effect during this period. The near end establishes a call in a conference room with high reverberation and tests the effect of reverberation reduction when different suppression degree parameters are set.

2. The noise elimination test method according to claim 1, characterized in that: The method for testing the intermediate audio information by using an audio test model comprises: Dividing the intermediate audio of the intermediate audio information into a plurality of intermediate speech frames; Calculating the speech energy of the intermediate speech frame by using an energy model, and smoothing the speech energy of the intermediate speech frame and the energy of the historical frame to obtain the energy value of the intermediate audio information; The energy model is: RMS i =αRMS cur +(1-α)RMS i-1 In the formula, the speech data of the i-th microphone is represented by x i1 , x i2 , …, x iL ; L is the speech frame length.

3. The noise elimination test method according to claim 1, characterized in that: The method further comprises: Respectively dividing the intermediate audio and the corresponding clean audio into a plurality of intermediate speech frames and clean speech frames, and deleting the silence frames therein to obtain intermediate data frames and clean data frames; Adding a Hanning window to the intermediate data frame and the clean data frame respectively and calculating the corresponding short-time DFT signal and 1 / 3 octave time-frequency decomposition signal; The 1 / 3 octave time-frequency decomposition signals corresponding to the intermediate data frame and the pure data frame are normalized, and the STOI value of the intermediate audio information is determined according to the obtained normalization factor, signal distortion ratio and relationship number.

4. The noise elimination test method according to claim 1, characterized in that: The method further comprises: Performing noise testing and reverberation testing on the intermediate audio information simultaneously, separately or successively; The noise elimination effect of the test audio is evaluated according to the noise test results and the reverberation test results.

5. The noise elimination test method according to claim 4, characterized in that: The method for performing a reverberation test on the intermediate audio comprises: After filtering the intermediate audio input with a 23-channel filter having an equivalent rectangular bandwidth, the instantaneous envelope of each frequency band output signal is obtained by a Hilbert transform method; The instantaneous envelope is windowed and subjected to DFT transformation to obtain a reverberation signal; After the reverberation signal is input into an 8-channel auditory modulation filter for processing, the speech reverberation modulation ratio of the intermediate audio is obtained according to the ratio of the energy of the first 4 frequency bands to the energy of the last 4 frequency bands, and the result of reverberation of the intermediate audio is evaluated according to the speech reverberation modulation ratio.

6. A noise cancellation test system, characterized in that: The system comprises: An acquisition module, used for acquiring initial audio information and performing preliminary processing on the initial audio information to obtain test audio information; A testing module, used for dividing the speech sequence of the test audio information into a plurality of test speech frames, and performing noise suppression and reverberation suppression processing on the test speech frames to obtain intermediate audio information; An evaluation module, used to test the intermediate audio information through an audio test model, and evaluate the noise elimination effect of the test audio according to the test result; The initial audio information is preliminarily processed to obtain test audio information, including: Determining the type of the current audio signal according to the initial audio information; When the type of the current audio signal is scene audio, testing the scene audio according to the target audio information, and using the scene audio that passes the test as the first test audio; When the type of the current audio signal is pure audio, reverberation processing is performed on the pure audio and the reference audio, and the delayed reverberation reference audio is mixed with the reverberation pure audio to obtain a second test audio; The testing the scene audio according to the target audio information includes: After the near end establishes a call, speak alternately with the far end normally to test the convergence speed of background noise suppression and the effect after convergence; After the near end establishes a call, keep talking to test the effect of noise cancellation when it has not converged; Speak in an environment with relatively loud but stable near-end background noise to test the noise cancellation effect during speaking; Speak in a near-end burst noise environment to test the noise cancellation effect during speaking; There is background voice, phone ringing, or music playing at the near end. Test the noise cancellation effect during this period. The near end establishes a call in a conference room with high reverberation and tests the effect of reverberation reduction when different suppression degree parameters are set.

7. An audio and video device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images