Data acquisition method and device, electronic equipment and storage medium

Abstract

The invention discloses a data acquisition method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring to-be-processed audios collected by a microphone array, wherein the to-be-processed audios comprise multiple sentences of multi-channel audios; and for each sentence of multi-channel audio, phase difference features between at least one group of specified channel audios are obtained as target data, and the specified channel audios comprise at least two channel audios with associated positions. According to the method, for each sentence of multi-channel audio, phase difference calculation is carried out through the audio including at least two microphone channels associated in position, on one hand, higher-dimension phase information coverage can be generated, DOA features obtained after phase difference are more dense, and therefore the accuracy of a DOA coding result can be improved; and on the other hand, the phase difference calculation is carried out by adopting the position-associated microphone channel audio, so that data redundancy can be avoided, and the DOA coding efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of data acquisition, and more specifically, to a data acquisition method, device, electronic device and storage medium. Background Art

[0002] Multi-channel microphone arrays can estimate the direction of arrival (DOA) of sound based on the target speech extraction (TSE) algorithm of neural networks, thereby estimating the direction and distance of the sound source. In the process of multi-channel microphone array operation, the DOA features need to be encoded. However, the DOA features obtained by the current phase difference between channels are relatively sparse, making it difficult to obtain accurate DOA encoding results. Summary of the invention

[0003] The present application proposes a data acquisition method, device, electronic device and storage medium to improve the above-mentioned problems.

[0004] In a first aspect, an embodiment of the present application provides a data acquisition method, the method comprising: acquiring audio to be processed collected by a microphone array, the audio to be processed comprising multiple sentences of multi-channel audio; for each sentence of multi-channel audio, acquiring a phase differential feature between at least one group of specified channel audios as target data, the specified channel audio comprising at least two channel audios that are positionally associated.

[0005] In a second aspect, an embodiment of the present application provides a data acquisition device, comprising: a module for acquiring audio to be processed, for acquiring audio to be processed collected by a microphone array, wherein the audio to be processed includes multiple sentences of multi-channel audio; a data acquisition module, for acquiring, for each sentence of multi-channel audio, a phase differential feature between at least one group of specified channel audios as target data, wherein the specified channel audio includes at least two channel audios that are positionally associated.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, comprising: one or more processors; a memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to execute the data acquisition method provided in the first aspect above.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a program code is stored. The program code can be called by a processor to execute the data acquisition method provided in the first aspect above.

[0008] The present application provides a data acquisition method, device, electronic device and storage medium, which acquires the audio to be processed collected by a microphone array, wherein the audio to be processed includes multiple sentences of multi-channel audio; for each sentence of multi-channel audio, the phase difference feature between at least one set of specified channel audio is acquired as the target data, wherein the specified channel audio includes at least two position-associated channel audio. Thus, for each sentence of multi-channel audio, by incorporating the audio of at least two position-associated microphone channels for phase difference calculation, on the one hand, a higher-dimensional phase information coverage can be generated, so that the DOA features obtained after phase difference calculation are more dense, thereby helping to improve the accuracy of the DOA encoding result; on the other hand, since the position-associated microphone channel audio is used for phase difference calculation, data redundancy can be avoided, thereby helping to improve the efficiency of DOA encoding. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flow chart of a data acquisition method provided by an embodiment of the present application is shown.

[0011] Figure 2 An example diagram of the shape of a microphone array provided in an embodiment of the present application is shown.

[0012] Figure 3 Shows Figure 1 Flow chart of the method of step S120.

[0013] Figure 4 A structural example diagram of a DOA encoder provided in an embodiment of the present application is shown.

[0014] Figure 5 A flow chart of a data acquisition method provided by another embodiment of the present application is shown.

[0015] Figure 6 An example diagram of position arrangement between different channel audios when the designated channel audio provided by an embodiment of the present application is four-channel audios is shown.

[0016] Figure 7 A structural block diagram of a data acquisition device provided in an embodiment of the present application is shown.

[0017] Figure 8 A structural block diagram of an electronic device provided in an embodiment of the present application is shown.

[0018] Fig. 9 A storage unit for storing or carrying program codes for implementing the data acquisition method according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0020] Directional sound pickup is one of the many methods of speech enhancement. It refers to picking up the target signal from the mixed signal according to the direction of the sound source. That is, only the sound signal propagating from a specific direction is picked up, while the noise and interference signals from other directions are not picked up but attenuated or shielded, thereby achieving the effect of target speech enhancement.

[0021] Multi-channel microphone arrays can estimate the direction of arrival (DOA) of sound based on the target speech extraction (TSE) algorithm of the neural network, thereby estimating the direction and distance of the sound source. In the process of multi-channel microphone array operation, the DOA features need to be encoded. However, the DOA features obtained by the current phase difference between channels are relatively sparse, making it difficult to obtain accurate DOA encoding results.

[0022] After long-term research, the inventors found that it is possible to obtain the audio to be processed collected by a microphone array, and the audio to be processed includes multiple sentences of multi-channel audio; for each sentence of multi-channel audio, obtain at least one set of phase difference features between specified channel audios as target data, and the specified channel audio includes at least two position-associated channel audios. Therefore, for each sentence of multi-channel audio, by incorporating the audio of at least two position-associated microphone channels for phase difference calculation, on the one hand, a higher-dimensional phase information coverage can be generated, making the DOA features obtained after phase difference more dense, thereby helping to improve the accuracy of the DOA encoding result; on the other hand, since the position-associated microphone channel audio is used for phase difference calculation, data redundancy can be avoided, thereby helping to improve the efficiency of DOA encoding.

[0023] Therefore, in order to improve the above-mentioned problems, the inventors proposed the data acquisition method, device, electronic device and storage medium provided in the present application, which can produce higher-dimensional phase information coverage, so that the DOA features obtained after phase differentiation are denser, thereby helping to improve the accuracy of the DOA coding results, and can avoid data redundancy, thereby helping to improve the efficiency of DOA coding.

[0024] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0025] See also Figure 1 , shows a flow chart of a data acquisition method provided by an embodiment of the present application, and this embodiment provides a data acquisition method that can be applied to electronic devices, wherein the electronic device in this embodiment can be a mobile communication device with a network connection function such as a smart phone, and the specific device type may not be limited. The method includes:

[0026] Step S110: Acquire the audio to be processed collected by the microphone array, where the audio to be processed includes multi-sentence multi-channel audio.

[0027] In the implementation manner of the present application, the microphone array may include multiple microphones, and the audio to be processed collected by the microphone array may include multiple sentences (which may be understood as several sentences, and the specific number may not be limited) of multi-channel audio, that is, each sentence of the audio to be processed is a multi-channel reverberation audio (the reverberation audio may include various environmental noises). The specific type of the audio to be processed may not be limited, for example, it may be user voice, singing, video audio, etc.

[0028] In this embodiment, the specific shape of the microphone array is not limited, for example, it can be a ring (circular) microphone array, a rectangular microphone array or other shaped microphone arrays. The embodiment of the present application is described by taking the microphone array as a circular microphone array as an example.

[0029] For example, see Figure 2 , shows an example diagram of the shape of the microphone array provided in the embodiment of the present application. Figure 2 As shown, the microphone array is a circular array, including 9 microphone channels, 8 of which correspond to the circular ring of the microphone array, and the 9th microphone channel is located at the center of the circular ring. For the convenience of description, it can be referred to as the center microphone. Optionally, the audio collected by the center microphone can be called the center channel audio. The audio collected by each microphone channel can be multi-channel reverberation audio.

[0030] Step S120: For each sentence of multi-channel audio, obtain a phase difference feature between at least one group of designated channel audios as target data, wherein the designated channel audios include at least two channel audios associated with each other in position.

[0031] Each sentence of multi-channel audio includes multiple channel audios, and the designated channel audio may include at least two channel audios that are positionally associated among the multiple channel audios. For example, assuming that a sentence of multi-channel audio includes 10 channel audios, the designated channel audio may be two channel audios that are positionally associated among the 10 channel audios, the designated channel audio may be three channel audios that are positionally associated among the 10 channel audios, or the designated channel audio may be four channel audios that are positionally associated among the 10 channel audios.

[0032] In this embodiment, the position association can be understood as adjacent positions or close positions. Figure 2 Taking the microphone array shown as an example, since the microphone array has a center microphone, in this way, the position association can be understood as the position correspondence between the audio collected by any microphone located on the ring and the center channel audio, or the position association can be understood as the position correspondence between the audio collected by any two adjacent microphones located on the ring (optionally, when the microphone array does not have a center microphone, the position association can also be determined in this way).

[0033] In order to reduce data redundancy as much as possible, phase difference processing can be performed on the audio to be processed. Specifically, for each sentence of multi-channel audio, a phase difference feature between at least one set of specified channel audios can be obtained as target data.

[0034] If you only need to focus on the audio in a specific direction, you can obtain the phase difference features between a group of specified channel audios corresponding to the specific direction. Figure 2 Taking the microphone array shown as an example, when the specified channel audio is two-channel audio, the phase difference characteristics between the audio collected by microphone 1 (not limited here, it can also be any one of 2, 3, 4, 5, 6, 7, 8) and microphone 9 can be obtained.

[0035] Optionally, if you need to focus on audio in multiple directions, you can obtain phase difference features between multiple groups of specified channel audio corresponding to multiple directions (which can be several specific directions or omnidirectional). Figure 2 Taking the microphone array shown as an example, when the designated channel audio is still two-channel audio, the phase differential characteristics between the audio collected by microphone 1 and microphone 9 can be obtained, the phase differential characteristics between the audio collected by microphone 2 and microphone 9 can be obtained, the phase differential characteristics between the audio collected by microphone 3 and microphone 9 can be obtained, the phase differential characteristics between the audio collected by microphone 4 and microphone 9 can be obtained, the phase differential characteristics between the audio collected by microphone 5 and microphone 9 can be obtained, the phase differential characteristics between the audio collected by microphone 6 and microphone 9 can be obtained, the phase differential characteristics between the audio collected by microphone 7 and microphone 9 can be obtained, and the phase differential characteristics between the audio collected by microphone 8 and microphone 9 can be obtained.

[0036] It should be noted that the above examples are only for illustration and do not constitute a limitation on the number of audio channels specified. Figure 2Taking the microphone array shown as an example, when the designated channel audio is still three-channel audio, if you need to focus on the audio in a specific direction, you can obtain the phase difference characteristics between the audio collected by microphone 1, microphone 9 and microphone 2; if you need to focus on omnidirectional audio, you can obtain the phase difference characteristics between the audio collected by microphone 1, microphone 9 and microphone 2, obtain the phase difference characteristics between the audio collected by microphone 2, microphone 9 and microphone 3, obtain the phase difference characteristics between the audio collected by microphone 3, microphone 9 and microphone 4, obtain the phase difference characteristics between the audio collected by microphone 4, microphone 9 and microphone 5, obtain the phase difference characteristics between the audio collected by microphone 5, microphone 9 and microphone 6, obtain the phase difference characteristics between the audio collected by microphone 6, microphone 9 and microphone 7, obtain the phase difference characteristics between the audio collected by microphone 7, microphone 9 and microphone 8, and obtain the phase difference characteristics between the audio collected by microphone 8, microphone 9 and microphone 1.

[0037] Optionally, for each sentence of multi-channel audio, a process of obtaining phase difference features between at least one set of specified channel audios is described as follows.

[0038] See also Figure 3 As an implementation manner, step S120 may specifically include:

[0039] Step S121: Perform Fourier transform on each sentence of multi-channel audio to obtain real part parameters and imaginary part parameters corresponding to each channel audio.

[0040] As an implementation method, each sentence of multi-channel audio may be Fourier transformed to obtain the real and imaginary parameters corresponding to each channel audio. The size of each sentence of multi-channel audio after Fourier transformation is [N, C, F, T], where N represents the number of training samples, C represents the number of channels, F represents the frequency dimension, and T represents the time dimension.

[0041] The specific principle and implementation process of Fourier transform will not be described here.

[0042] Step S122: Based on the corresponding real part parameter and imaginary part parameter, a first phase difference and a second phase difference between at least one set of designated channel audios are obtained, wherein the second phase difference differs from the first phase difference by a sine of ninety degrees.

[0043] As an implementation manner, a first phase difference and a second phase difference between at least one set of specified channel audios may be obtained based on the corresponding real and imaginary parameters. In this implementation manner, the second phase difference differs from the first phase difference by a sine of ninety degrees.

[0044] As an implementation manner, when the designated channel audio is two-channel audio, the first phase difference and the second phase difference between at least one group of designated channel audio may be obtained according to the following formula:

[0045] IPD(W k1 ,W k2 )={cos(arctan2(R k1 ,I k1 )-arctan2(R k2 ,I k2 )),sin(arctan2(R k1 ,I k1 )-arctan2(R k2 ,I k2 ))}

[0046] Among them, cos(arctan2(R k1 ,I k1 )-arctan2(R k2 ,I k2 )) represents the first phase difference, sin(arctan2(R k1 ,I k1 )-arctan2(R k2 ,I k2 )) represents the second phase difference, R k1 , R k2 Characterizing the real part parameter, I k1 ,I k2 Characterizes the imaginary part parameter, at this time IPD (W k1 ,W k2 ) characterizes the phase difference parameter between specified channels.

[0047] Optional, IPD(W k1 ,W k2 ) is (N, C IPD , 2*F,T), where N represents the number of training samples, C IPD It represents the number of pairs of channels calculated, F represents the frequency dimension, and T represents the time dimension.

[0048] Optionally, in the case where the designated channel audio is two-channel audio, the designated channel audio may be the two-channel audio associated with the aforementioned positions, or may be two-channel audio selected according to a designated rule.

[0049] Optionally, the designated rule may be to select the audio collected by the two microphone channels with the closest physical distance in the microphone array as the designated channel audio, or to select the audio collected by the two microphone channels with the farthest physical distance in the microphone array as the designated channel audio, or to select the audio collected by any two microphone channels in the microphone array as the designated channel audio (in this case, all microphones in the microphone array will form a pair of two, and each pair of microphones will serve as a designated channel). For example, when the microphone array is Figure 2 In the case of the circular array shown, each audio sentence includes 9 microphone channels, which are denoted as W for ease of description. k = {W k1 ,W k2 ,…,W k9}, at this time, each of the 8 microphones located on the ring can be combined with the central microphone to form a designated channel in pairs. In this way, 8 groups of designated channels can be obtained, that is, C=8.

[0050] Step S123: concatenate the first phase difference and the second phase difference to obtain the phase difference feature as target data.

[0051] In this embodiment, in order to avoid data redundancy and reduce the subsequent data processing pressure, the first phase difference and the second phase difference obtained above can be spliced ​​to obtain a phase difference feature as the target data. The size of the phase difference feature obtained after splicing the first phase difference and the second phase difference is still (N, C IPD , 2*F,T).

[0052] In this implementation, the acquired target data may be used for DOA feature encoding. Specifically, the target data may be input into a DOA (Direction of arrival) encoder, and the DOA encoder may first encode the target data and then perform compression (dimensionality reduction) processing.

[0053] The DOA encoder may include a first coding layer, a second coding layer, and a third coding layer, and a bottleneck structure is formed between the first coding layer, the second coding layer, and the third coding layer (eg Figure 4 As shown), the output of the first coding layer enters the second coding layer, and the output of the second coding layer enters the third coding layer.

[0054] Optionally, the target data in this embodiment may include the number of channels and the frequency dimension, and each sentence of the audio to be processed includes a label, which can be used to guide the model to learn the audio processing operations required for different pickup areas. In this way, in the process of reducing the number of channels and the frequency dimension, as an implementation method, the number of channels can be reduced by the first dimension and the frequency dimension can be reduced by the second dimension in the first coding layer of the DOA encoder; then, in the second coding layer of the DOA encoder, the frequency dimension output by the first coding layer is reduced by the second dimension, and the dimension of the number of channels output by the first coding layer is maintained, that is, the second coding layer may not perform the dimension reduction process on the number of channels, but only transmit the number of channels output by the first coding layer to the third coding layer; then, in the third coding layer of the DOA encoder, the dimension of the number of channels transmitted by the second coding layer is increased by the first dimension, and the frequency dimension output by the second coding layer is reduced by the second dimension.

[0055] The first dimension is determined based on the number of channel audio pairs consisting of two pairs of specified channel audio, and the second dimension is determined based on the number of sampling points of Fourier transform. Optionally, the number of channel audio pairs consisting of two pairs of specified channel audio is an even number of pairs.

[0056] In some other implementations, the number of channel audio pairs consisting of pairs of channel audios of the specified channel audios may also be an odd number of pairs. In this case, the initial frequency dimension is F. If the number of channel audio pairs consisting of pairs of channel audios of the specified channel audios is an even number of pairs, the initial frequency dimension may be 2F.

[0057] Optionally, in a specific embodiment, the number of channels can be reduced by 4 times (first dimension) and the frequency dimension can be reduced by 2 times (second dimension) in the first coding layer; the frequency dimension of the output of the first coding layer can be reduced by 2 times (second dimension) in the second coding layer, and the dimension of the number of channels output by the first coding layer can be maintained; and in the third coding layer, the number of channels transmitted by the second coding layer can be increased by 4 times (first dimension), and the frequency dimension of the output of the second coding layer can be further reduced by 2 times (second dimension).

[0058] Among them, the frequency dimension is a redundant part in the DOA encoder, so it can be reduced in dimension multiple times. After the channel number is reduced in dimension, it will be increased to the original dimension in order to facilitate the complete output of the channel characteristics of the audio.

[0059] In the implementation manner of the present application, the second coding layer can be expanded to n layers (n≥1), and optionally, each of the n layers also only performs dimensionality reduction processing on the frequency dimension. Optionally, each layer can reduce the frequency dimension to half of the previous layer.

[0060] In the implementation manner of the present application, the number of channels output by the third coding layer may be 4 or an even number divisible by 4. Optionally, the number of channels output by the third coding layer may be as low as 4. Among them, the 4 channels can enhance modeling of four different right-angle directions in the plane. For example, by connecting two microphones with a line and intersecting the two lines, a coordinate system can be constructed, and the coordinate system can be used to enhance modeling of four different directions in the plane.

[0061] In this implementation manner, the number of channels and the frequency dimension output by the third coding layer of the DOA encoder may be used as output parameters of the DOA encoder.

[0062] Furthermore, the output parameters of the DOA encoder may be input into a specified network layer of the neural network to be trained, so as to train the neural network to be trained starting from the specified network layer to obtain a target neural network.

[0063] Among them, the specified network layer is the hidden layer at the specified position, and the specified position can be understood as the network layer corresponding to the third coding layer position of the DOA encoder in the neural network to be trained. For example, the specified position can be the third network layer or the fourth network layer of the neural network to be trained, and the third network layer and the fourth network layer are both hidden layers.

[0064] Among them, when the output parameters of the DOA encoder are input into the specified network layer of the neural network to be trained, the output parameters of the DOA encoder can be spliced ​​with the input features of the specified network layer to obtain reference input features, and the reference input features are used as new input features of the specified network layer. All hidden layers of the neural network to be trained starting from the current hidden layer are updated, and then the updated neural network is used as the target neural network.

[0065] Furthermore, the target neural network can be used to perform noise reduction processing on the processed audio. The target neural network trained by the implementation method of the present application has the ability to perform different audio processing on the audio of different pickup areas. Among them, the audio can come from the pickup area or the non-pickup area. If the audio comes from the pickup area, the target neural network model can perform speech signal enhancement processing on the audio according to the label carried by the sample audio (the label at this time is the sample audio or the reverberation audio); and if the audio comes from the non-pickup area, the target neural network model can perform suppression processing on the audio according to the label carried by the sample audio (the label at this time is silent audio).

[0066] As an implementation method, the target neural network can be used to perform noise reduction on the audio to be processed, that is, the target neural network can remove the reverberation sound (which can also be understood as reflected sound) in the audio to be processed, thereby achieving the effects of noise reduction, reverberation removal and directional sound pickup. The audio to be processed can be an audio of the same type as the sample audio but with different content.

[0067] The data acquisition method provided in this embodiment obtains the audio to be processed collected by the microphone array, and the audio to be processed includes multiple sentences of multi-channel audio; for each sentence of multi-channel audio, the phase difference feature between at least one group of specified channel audio is obtained as the target data, and the specified channel audio includes at least two channel audios associated with positions. Therefore, for each sentence of multi-channel audio, by incorporating the audio of at least two microphone channels associated with positions to perform phase difference calculation, on the one hand, a higher-dimensional phase information coverage can be generated, so that the DOA features obtained after phase difference calculation are more dense, thereby helping to improve the accuracy of the DOA encoding result; on the other hand, since the phase difference calculation is performed using microphone channel audio associated with positions, data redundancy can be avoided, thereby helping to improve the efficiency of DOA encoding.

[0068] See also Figure 5 , shows a flow chart of a data acquisition method provided by another embodiment of the present application. This embodiment provides a data acquisition method that can be applied to an electronic device. The method includes:

[0069] Step S210: Acquire the audio to be processed collected by the microphone array, where the audio to be processed includes multi-sentence multi-channel audio.

[0070] The specific implementation of step S210 may refer to the relevant description of step S110 in the aforementioned embodiment, which will not be repeated here.

[0071] Step S220: Perform Fourier transform on each sentence of multi-channel audio to obtain real part parameters and imaginary part parameters corresponding to each channel audio.

[0072] The specific implementation of step S220 may refer to the relevant description of step S121 in the aforementioned embodiment, which will not be repeated here.

[0073] Step S230: Based on the corresponding real part parameter and imaginary part parameter, a first phase difference and a second phase difference between at least one set of designated channel audios are obtained, wherein the second phase difference differs from the first phase difference by a sine of ninety degrees.

[0074] Part of the specific implementation of step S230 may refer to the relevant description of step S122 in the aforementioned embodiment, which will not be repeated here.

[0075] In this implementation manner, the number of designated channel audios is different, and the manner of obtaining the first phase difference and the second phase difference between at least one group of designated channel audios may be different. As an implementation manner, in the process of obtaining the first phase difference and the second phase difference between at least one group of designated channel audios based on the corresponding real parameter and the imaginary parameter, the number of channel audios included in the designated channel audios may be first determined, and then the first phase difference and the second phase difference between at least one group of the number of channel audios may be obtained based on the corresponding real parameter and the imaginary parameter.

[0076] Optionally, the number of channel audios included in the designated channel audios in this implementation may not be specifically limited. The implementation of the present application is described by taking the example that the designated channel audios include two channel audios, three channel audios, and four channel audios respectively.

[0077] Among them, when the designated channel audio includes two channel audios (which can be understood as first-order channel audios), if the audio to be processed collected by the microphone array includes the center channel audio, the designated channel audio can include the center channel audio and any other channel audio. In this way, if the number of designated channel audios is a group, the phase difference between any other channel audio and the center channel audio can be performed based on the real parameters and imaginary parameters corresponding to any other channel audio and the real parameters and imaginary parameters corresponding to the center channel audio to obtain a phase difference result; then the cosine value of the phase difference result is obtained as the first phase difference; and the sine value of the phase difference result is obtained as the second phase difference.

[0078] Optionally, if the audio to be processed collected by the microphone array does not include the center channel audio, the specified channel audio may include any two adjacent channel audios. In this way, the two adjacent channel audios will be phase-differentiated based on the real parameters and imaginary parameters corresponding to each of the two adjacent channel audios to obtain a phase difference result.

[0079] Optionally, if the audio to be processed collected by the microphone array does not include the center channel audio, the specified channel audio may also include two channel audios at any position. In this way, the two channel audios at any position can be phase-differentiated based on the real parameters and imaginary parameters corresponding to the two channel audios at the arbitrary position to obtain a phase difference result.

[0080] In the case where the designated channel audio includes two channel audios, as an implementation mode, the phase difference feature may be obtained according to the following formula:

[0081] IPD(W k1 ,W k2)=concat{cos(arctan2(R k1 ,I k1 )-arctan2(R k2 ,I k2 )),sin(arctan2(R k1 ,I k1 )-arctan2(R k2 ,I k2 ))}.

[0082] Among them, IPD(W k1 ,W k2 ) represents the phase difference feature, and concat represents the concatenation of the first phase difference and the second phase difference along the F (frequency) axis.

[0083] Optionally, the above formula only cites one group (i.e., the case of 1-channel audio and 2-channel audio) for illustration. Figure 2 Taking the 9-channel circular microphone array shown as an example, the audio collected by the central channel of the microphone array can be used as the central channel audio of the specified channel audio. In this way, the following 8 groups of phase difference features (abbreviated as IPD features) can be constructed:

[0084] {W k1 ,W k9},{W k2 ,W k9},{W k3 ,W k9},{W k4 ,W k9},{W k5 ,W k9},{W k6 ,W k9},{W k7 ,W k9},{W k8 ,W k9}.

[0085] In this embodiment, when the designated channel audio includes three channel audios (which can be understood as second-order channel audios), if the audio to be processed collected by the microphone array includes the center channel audio, the designated channel audio may include the center channel audio and any two other channel audios adjacent to each other. In this way, if the number of designated channel audios is a group, each channel audio of two adjacent channel audios can be phase-differentiated with the center channel audio based on the real parameters and imaginary parameters corresponding to any two other adjacent channel audios and the real parameters and imaginary parameters corresponding to the center channel audio to obtain two phase difference results to be processed; and the two phase difference results to be processed are subtracted to obtain a target phase difference result; then the cosine value of the target phase difference result is obtained as the first phase difference; and the sine value of the target phase difference result is obtained as the second phase difference.

[0086] That is to say, this method will perform phase difference on the result of phase difference between each channel audio of two adjacent channel audios and the center channel audio, so as to avoid data redundancy caused by data overlap or intersection.

[0087] Optionally, if the audio to be processed collected by the microphone array does not include the center channel audio, the designated channel audio may include any three adjacent channel audios. In this manner, the middle channel audio of the three adjacent channel audios may be used as the center channel audio of the designated channel audio. In this manner, the method for calculating the first phase difference and the second phase difference is the same as the method for calculating when the audio to be processed collected by the microphone array includes the center channel audio, and is not described in detail here.

[0088] Optionally, if the audio to be processed collected by the microphone array does not include the center channel audio, the specified channel audio may also include three channel audios at any position. In this case, the method for calculating the first phase difference and the second phase difference is the same as the method for calculating when the audio to be processed collected by the microphone array includes the center channel audio. For example, assuming that the three channel audios are 1-channel audio, 2-channel audio, and 3-channel audio, the phase difference between the 1-channel audio and the 2-channel audio will be calculated, recorded as the first phase difference to be processed, the phase difference between the 2-channel audio and the 3-channel audio will be calculated, recorded as the second phase difference to be processed, and then the phase difference between the first phase difference to be processed and the second phase difference to be processed will be calculated to obtain the target phase difference result.

[0089] In the case where the specified channel audio includes three channel audios, as an implementation mode, the phase difference feature may be obtained according to the following formula:

[0090] IPD(W k1 ,W k2 ,W k3)=concat{cos(arctan2(R k1 ,I k1 )-2*arctan2(R k2 ,I k2 )+arctan2(R k3 ,I k3 )),

[0091] sin(arctan2(R k1 ,I k1 )-2*arctan2(R k2 ,I k2 )+arctan2(R k3 ,I k3 ))}

[0092] Among them, cos(arctan2(R k1 ,I k1 )-2*arctan2(R k2 ,I k2 )+arctan2(R k3 ,I k3 )) represents the first phase difference, sin(arctan2(R k1 ,I k1 )-2*arctan2(R k2 ,I k2 )+arctan2(R k3 ,I k3 )) characterizes the second phase difference, IPD(W k1 ,W k2 ,W k3 ) represents the phase difference feature, and concat represents the concatenation of the first phase difference and the second phase difference along the F (frequency) axis.

[0093] Optionally, the above formula only cites one group (i.e., 1-channel audio, 2-channel audio, and 3-channel audio) for illustration, and the 2-channel audio is used as the center channel audio of the designated channel audio. Figure 2 Taking the 9-channel circular microphone array shown in the figure as an example, the audio collected by the central channel of the microphone array can be used as the central channel audio of the specified channel audio. In this way, the following 8 groups of phase difference features (which can be referred to as hoIPD (high-order Inter-microphone phase difference, high-order inter-channel phase difference) features) can be constructed:

[0094] {W k1 ,W k9 ,W k2},{Wk2 ,W k9 ,W k3},{W k3 ,W k9 ,W k4},{W k4 ,W k9 ,W k5},{W k5 ,W k9 ,W k6},{W k6 ,W k9 ,W k7},{W k7 ,W k9 ,W k8},{W k8 ,W k9 ,W k1}

[0095] In this embodiment, when the designated channel audio includes four channel audios (which can be understood as third-order channel audios), if the microphone array includes two center channel audios (for example, when the microphone array is a concentric circle array layout, there will be two center channel audios), then the designated channel audio can include the two center channel audios and any two other channel audios adjacent to each other. In this way, if the number of designated channel audios is a group, each of the two adjacent channel audios can be phase-differentiated with a center channel audio with a closer interval based on the real parameters and imaginary parameters corresponding to each of the other two adjacent channel audios and the real parameters and imaginary parameters corresponding to each of the two center channel audios to obtain a first phase difference result to be processed and a second phase difference result to be processed; and based on the real parameters and imaginary parameters corresponding to each of the two center channel audios, the two center channel audios are phase-differentiated to obtain a third phase difference result to be processed; then the first phase difference result to be processed is subtracted from the third phase difference result to be processed to obtain a first reference phase difference result; and the second phase difference result to be processed is subtracted from the third phase difference result to be processed to obtain a second reference phase difference result; then the first reference phase difference result is subtracted from the second reference phase difference result to obtain a target phase difference result; then the cosine value of the target phase difference result is obtained as the first phase difference; and the sine value of the target phase difference result is obtained as the second phase difference.

[0096] That is, assuming that the four channel audios are 1-channel audio, 2-channel audio, 3-channel audio, and 4-channel audio, and the positions of the 1-channel audio, 2-channel audio, 3-channel audio, and 4-channel audio are arranged as follows: Figure 6 As shown, Figure 6 The two circles shown represent microphone arrays. The phase difference between 1-channel audio and 2-channel audio is first calculated, recorded as the first phase difference result to be processed; the phase difference between 3-channel audio and 4-channel audio is calculated, recorded as the second phase difference result to be processed; and the phase difference between 2-channel audio and 3-channel audio is calculated, recorded as the third phase difference result to be processed; then the phase difference result between the first phase difference result to be processed and the third phase difference result to be processed is calculated, recorded as the first reference phase difference result, and the phase difference result between the second phase difference result to be processed and the third phase difference result to be processed is calculated, recorded as the second reference phase difference result; then the phase difference result between the first reference phase difference result and the second reference phase difference result is calculated to obtain the target phase difference result.

[0097] By calculating multiple phase difference results, the association between different audio channels with similar positions can be established, so that the phase information of multiple different microphones can be effectively integrated at the feature level. In addition, by performing phase difference on the phase difference results multiple times, data redundancy can be reduced, which helps to reduce the learning pressure of subsequent DOA encoders and neural networks.

[0098] Optionally, if the microphone array does not include (two) center channel audios, the designated channel audio may include any four adjacent channel audios, in which case the middle two channel audios of the four adjacent channel audios may be used as the center channel audios of the designated channel audio. In this manner, the manner of calculating the first phase difference and the second phase difference is the same as the manner of calculating when the microphone array includes two center channel audios, and is not described in detail here.

[0099] Optionally, if the microphone array does not include (two) center channel audio, the specified channel audio may also include four channel audios at any position. In this case, the method of calculating the first phase difference and the second phase difference is the same as the method of calculating when the microphone array includes two center channel audios.

[0100] In the case where the specified channel audio includes four channel audios, as an implementation, the phase difference feature may be obtained according to the following formula:

[0101] IPD(W k1 , W k2 , W k3 , W k4 )=concat{cos((artan2(R k1 , I k1 )-arctan2(R k2 , I k2 ))-2*(arctan2(R k2 , Ik2 ) - arctan2(R k3 , I k3 )) + (arctan2(R k2 , I k3 ) - arctan2(R k4 , I k4 ))),sin((arctan2(R k1 , I k1 ) - arctan2(R k2 , I k2 ) - 22 * (arctan2(R k2 , I k2 ) - arctan2(R k3 , I k3 )) + (arctan2(R k3 , I k3 ) - arctan2(R k4 , I k4 )))} = concat{cos(arctan2(R k1 , I k1 ) - 3 * arctan2(R k2 , I k2 + 3 * arctan2(R k3 , I k3 ) - arctan2(R k4 , k4 )),sin(arctan2(R k1 , I k1 ) - 3 * arctan2(R k2 , I k2 ) + 3 * arctan2(R k3 , I k3 ) - arctan2(R k4 , I k4 ))}

[0102] 其中,cos((arctan2(R k1 , I k1 ) - arctan2(R k2 , I k2 )) - 2 * (arctan2(R k2 , I k2 ) - arctan2(R k3 , I k3 )) + (arctan2(R k3 , I k3 ) - arctan2(R k4 , I k4 )))

[0103] Characterize the first phase difference, sin((arctan2(R k1 , I k1 )-arctan2(R k2 , I k2 ))-2*(arctan2(R k2 , I k2 )-arctan2(R k3 , I k3 ))+(arctan2(R k3 , I k3 )-arctan2(R k4 , I k4 )))

[0104] Characterize the second phase difference, IPD(W k1 ,W k2 ,W k3 ,W k4 ) represents the phase difference feature, and concat represents the concatenation of the first phase difference and the second phase difference along the F (frequency) axis.

[0105] Optionally, the above formula only cites one group (i.e., 1-channel audio, 2-channel audio, 3-channel audio, and 4-channel audio) for illustration, and the 2-channel audio and 3-channel audio are used as the center channel audio of the specified channel audio, wherein the 2-channel audio and the 3-channel audio can be the two center channel audios of a concentric circle-shaped microphone array, or can be the two center channel audios located between the 1-channel audio and the 4-channel audio, and no limitation is made here.

[0106] It is worth noting that if data redundancy needs to be reduced, the implementation method of the present application may give priority to at least two position-associated audio channels as designated audio channels. This method enables the DOA encoder to more efficiently learn the differences between different channel audios based on the structural relationship between different channel audios; and if the structural relationship between different channel audios is not considered when selecting the designated channel audio, at least two channel audios can be arbitrarily selected as the designated channel audio. This method allows the DOA encoder to more comprehensively learn the differences between different channel audios based on the structural relationship between different channel audios.

[0107] Step S240: concatenate the first phase difference and the second phase difference to obtain the phase difference feature as target data, wherein the designated channel audio includes at least two channel audios that are positionally associated.

[0108] The specific implementation of step S240 may refer to the relevant description of step S123 in the aforementioned embodiment, which will not be repeated here.

[0109] The data acquisition method provided in this embodiment obtains the audio to be processed collected by the microphone array, wherein the audio to be processed includes multiple sentences of multi-channel audio; performs Fourier transform on each sentence of the multi-channel audio to obtain the real part parameter and the imaginary part parameter corresponding to each channel audio; obtains the first phase difference and the second phase difference between at least one group of specified channel audios based on the corresponding real part parameter and the imaginary part parameter, wherein the second phase difference differs from the first phase difference by ninety degrees sinusoidally; splices the first phase difference and the second phase difference to obtain the phase difference feature as the target data, wherein the specified channel audio includes at least two position-associated channel audios. Thus, for each sentence of multi-channel audio, by incorporating the audio of at least two position-associated microphone channels for phase difference calculation, on the one hand, a higher-dimensional phase information coverage can be generated, so that the DOA feature obtained after phase difference calculation is more dense, thereby helping to improve the accuracy of the DOA encoding result; on the other hand, since the position-associated microphone channel audio is used for phase difference calculation, data redundancy can be avoided, thereby helping to improve the efficiency of DOA encoding.

[0110] At the same time, by introducing the phase difference algorithm between high-order (i.e. second-order and third-order) channel audios, the phase information of multiple different microphones can be effectively integrated at the feature level, thereby expanding the time delay information; and the IPD features obtained after the phase difference between the high-order channel audios can enable the DOA encoder to learn more comprehensively, thereby reducing the learning pressure on the DOA encoder and subsequent neural networks.

[0111] See also Figure 7 , is a structural block diagram of a data acquisition device provided in an embodiment of the present application. This embodiment provides a data acquisition device 300, which can be run on an electronic device. The device 300 includes: a to-be-processed audio acquisition module 310 and a data acquisition module 320:

[0112] The to-be-processed audio acquisition module 310 is used to acquire the to-be-processed audio collected by the microphone array, where the to-be-processed audio includes multi-sentence multi-channel audio.

[0113] The data acquisition module 320 is used to acquire, for each sentence of multi-channel audio, a phase difference feature between at least one group of designated channel audios as target data, wherein the designated channel audios include at least two channel audios associated with each other in position.

[0114] As an implementation method, the data acquisition module 320 can be used to perform Fourier transform on each sentence of multi-channel audio to obtain the real parameters and imaginary parameters corresponding to each channel audio; based on the corresponding real parameters and imaginary parameters, obtain the first phase difference and the second phase difference between at least one group of specified channel audios, the second phase difference being ninety degrees sinusoidally different from the first phase difference; splice the first phase difference and the second phase difference to obtain the phase differential feature as the target data.

[0115] In this embodiment, if the audio to be processed collected by the microphone array includes the center channel audio, the designated channel audio includes the center channel audio and any other channel audio. In this way, if the number of the designated channel audio is a group, the data acquisition module 320 can be specifically used to perform phase difference between the any other channel audio and the center channel audio based on the real parameters and imaginary parameters corresponding to the any other channel audio and the real parameters and imaginary parameters corresponding to the center channel audio to obtain a phase difference result; obtain the cosine value of the phase difference result as the first phase difference; obtain the sine value of the phase difference result as the second phase difference.

[0116] If the audio to be processed collected by the microphone array does not include the center channel audio, the designated channel audio includes any two adjacent channel audios.

[0117] In this embodiment, if the audio to be processed collected by the microphone array includes the center channel audio, the designated channel audio includes the center channel audio and any two adjacent channel audios. In this way, if the number of the designated channel audios is a group, the data acquisition module 320 can be specifically used to perform phase difference between each channel audio of two adjacent channel audios and the center channel audio based on the real parameters and imaginary parameters corresponding to the other two adjacent channel audios and the real parameters and imaginary parameters corresponding to the center channel audio to obtain two phase difference results to be processed; perform a difference on the two phase difference results to be processed to obtain a target phase difference result; obtain the cosine value of the target phase difference result as the first phase difference; obtain the sine value of the target phase difference result as the second phase difference.

[0118] Among them, if the audio to be processed collected by the microphone array does not include the center channel audio, the designated channel audio includes any three adjacent channel audios. In this way, the middle channel audio of the three adjacent channel audios can be used as the center channel audio of the designated channel audio.

[0119] In this embodiment, if the microphone array includes two center channel audios, the designated channel audio includes the two center channel audios and any two other channel audios adjacent to each other. In this way, if the number of the designated channel audios is a group, the data acquisition module 320 can be specifically used to perform phase difference between each channel audio of the two adjacent channel audios and a center channel audio with a closer interval based on the real parameters and imaginary parameters corresponding to the any two other channel audios adjacent to each other and the real parameters and imaginary parameters corresponding to the two center channel audios, to obtain a first to-be-processed phase difference result and a second to-be-processed phase difference result; based on the real parameters and imaginary parameters corresponding to the two center channel audios, the two center channel audios are phase-differentiated. The first phase difference result to be processed is subjected to phase difference to obtain a third phase difference result to be processed; the first phase difference result to be processed is subtracted from the third phase difference result to be processed to obtain a first reference phase difference result; the second phase difference result to be processed is subtracted from the third phase difference result to be processed to obtain a second reference phase difference result; the first reference phase difference result is subtracted from the second reference phase difference result to obtain a target phase difference result; the cosine value of the target phase difference result is obtained as the first phase difference; the sine value of the target phase difference result is obtained as the second phase difference.

[0120] Among them, if the audio to be processed collected by the microphone array does not include the center channel audio, the designated channel audio includes any four adjacent channel audios. In this way, the middle two channel audios of the four adjacent channel audios can be used as the center channel audio of the designated channel audio.

[0121] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0122] In several embodiments provided in the present application, the coupling between modules may be electrical, mechanical or other forms of coupling.

[0123] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0124] See also Figure 8Based on the above data acquisition method and device, the embodiment of the present application also provides an electronic device 100 that can execute the above data acquisition method. The electronic device 100 includes a memory 102 and one or more (only one is shown in the figure) processors 104 coupled to each other, and the memory 102 and the processor 104 are connected by a communication line. The memory 102 stores a program that can execute the content of the above embodiments, and the processor 104 can execute the program stored in the memory 102.

[0125] Among them, the processor 104 may include one or more processing cores. The processor 104 uses various interfaces and lines to connect various parts of the entire electronic device 100, and executes various functions and processes data of the electronic device 100 by running or executing instructions, programs, code sets or instruction sets stored in the memory 102, and calling data stored in the memory 102. Optionally, the processor 104 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 104 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing display content; and the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 104, but may be implemented separately through a communication chip.

[0126] The memory 102 may include a random access memory (RAM) or a read-only memory (ROM). The memory 102 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 102 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the aforementioned embodiments, etc. The data storage area may also store data (such as a phone book, audio and video data, chat record data) created by the electronic device 100 during use.

[0127] Please refer to Fig. 9, which shows a structural block diagram of a computer-readable storage medium provided in an embodiment of the present application. The computer-readable storage medium 400 stores program codes, which can be called by a processor to execute the method described in the above method embodiment.

[0128] The computer readable storage medium 400 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium 400 includes a non-transitory computer-readable storage medium. The computer readable storage medium 400 has storage space for program code 410 that performs any method step of the above method. These program codes can be read from or written to one or more computer program products. The program code 410 can be compressed, for example, in an appropriate form.

[0129] In summary, the embodiments of the present application provide a data acquisition method, device, electronic device and storage medium, which acquires the audio to be processed collected by a microphone array, wherein the audio to be processed includes multiple sentences of multi-channel audio; for each sentence of multi-channel audio, the phase difference feature between at least one group of specified channel audios is acquired as target data, wherein the specified channel audio includes at least two channel audios associated with positions. Thus, for each sentence of multi-channel audio, by incorporating the audio of at least two microphone channels associated with positions for phase difference calculation, on the one hand, a higher dimensional phase information coverage can be generated, so that the DOA features obtained after phase difference calculation are denser, thereby helping to improve the accuracy of the DOA encoding result; on the other hand, since the phase difference calculation is performed using the microphone channel audio associated with positions, data redundancy can be avoided, thereby helping to improve the efficiency of DOA encoding.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A data acquisition method, characterized in that: The method comprises: Acquire the audio to be processed collected by the microphone array, wherein the audio to be processed includes multiple sentences and multi-channel audio; For each sentence of multi-channel audio, a phase difference feature between at least one group of specified channel audios is obtained as target data, where the specified channel audios include at least two channel audios associated with each other in position.

2. The method according to claim 1, characterized in that The method of obtaining, for each sentence of multi-channel audio, at least one set of phase difference features between the specified channel audios as target data includes: Performing Fourier transform on each sentence of multi-channel audio to obtain real part parameters and imaginary part parameters corresponding to each channel audio; Based on the corresponding real part parameter and imaginary part parameter, obtaining a first phase difference and a second phase difference between at least one set of specified channel audio, wherein the second phase difference differs from the first phase difference by a sine of ninety degrees; The first phase difference and the second phase difference are concatenated to obtain the phase difference feature as target data.

3. The method according to claim 2, characterized in that The obtaining, based on the corresponding real part parameter and imaginary part parameter, a first phase difference and a second phase difference between at least one set of specified channel audios, comprises: Determine the number of channel audios included in the specified channel audio; Based on the corresponding real part parameters and imaginary part parameters, a first phase difference and a second phase difference between at least one group of the number of channel audios are obtained.

4. The method according to claim 3, characterized in that If the number of the designated channel audios is a group, and if the audio to be processed collected by the microphone array includes a center channel audio, the designated channel audio includes the center channel audio and any other channel audio, and the obtaining of a first phase difference and a second phase difference between at least one group of the number of channel audios based on the corresponding real parameter and imaginary parameter includes: Based on the real parameter and the imaginary parameter corresponding to any one of the other channel audios and the real parameter and the imaginary parameter corresponding to the center channel audio, performing phase difference on the any one of the other channel audios and the center channel audio to obtain a phase difference result; Obtaining a cosine value of the phase difference result as the first phase difference; A sine value of the phase difference result is obtained as the second phase difference.

5. The method according to claim 4, characterized in that If the audio to be processed collected by the microphone array does not include center channel audio, the designated channel audio includes any two adjacent channel audios.

6. The method according to claim 3, characterized in that If the number of the designated channel audios is a group, and if the audio to be processed collected by the microphone array includes a center channel audio, the designated channel audio includes the center channel audio and any two other channel audios adjacent to each other, the obtaining of a first phase difference and a second phase difference between at least one group of the number of channel audios based on the corresponding real part parameters and imaginary part parameters includes: Based on the real parameters and imaginary parameters corresponding to the two other adjacent channel audios and the real parameters and imaginary parameters corresponding to the center channel audio, phase-differentiate each channel audio of the two adjacent channel audios with the center channel audio to obtain two phase difference results to be processed; Subtracting the two phase difference results to be processed to obtain a target phase difference result; Obtaining a cosine value of the target phase difference result as the first phase difference; The sine value of the target phase difference result is obtained as the second phase difference.

7. The method according to claim 6, characterized in that If the audio to be processed collected by the microphone array does not include the center channel audio, and the designated channel audio includes any three adjacent channel audios, the method further includes: The middle channel audio of the three adjacent channel audios is used as the central channel audio of the designated channel audio.

8. The method according to claim 3, characterized in that If the number of the designated channel audio is a group, and if the microphone array includes two center channel audios, the designated channel audio includes the two center channel audios and any two other channel audios adjacent to each other, and based on the corresponding real part parameters and imaginary part parameters, obtaining a first phase difference and a second phase difference between at least one group of the number of channel audios includes: Based on the real parameters and imaginary parameters corresponding to the two other channel audios at adjacent positions, and the real parameters and imaginary parameters corresponding to the two center channel audios, phase-differentiate each channel audio of the two adjacent channel audios with a center channel audio with a closer interval, to obtain a first to-be-processed phase difference result and a second to-be-processed phase difference result; Based on the real part parameters and the imaginary part parameters corresponding to the two center channel audios, respectively, phase-differentiate the two center channel audios to obtain a third to-be-processed phase difference result; Subtracting the first to-be-processed phase difference result from the third to-be-processed phase difference result to obtain a first reference phase difference result; Subtracting the second to-be-processed phase difference result from the third to-be-processed phase difference result to obtain a second reference phase difference result; Subtracting the first reference phase differential result from the second reference phase differential result to obtain a target phase differential result; Obtaining a cosine value of the target phase difference result as the first phase difference; The sine value of the target phase difference result is obtained as the second phase difference.

9. The method according to claim 8, characterized in that If the audio to be processed collected by the microphone array does not include the center channel audio, and the designated channel audio includes any four adjacent channel audios, the method further includes: The middle two channel audios of the four adjacent channel audios are used as the central channel audios of the designated channel audio.

10. A data acquisition device, characterized in that: The device comprises: The to-be-processed audio acquisition module is used to acquire the to-be-processed audio collected by the microphone array, wherein the to-be-processed audio includes multi-sentence multi-channel audio; The data acquisition module is used to acquire, for each sentence of multi-channel audio, a phase difference feature between at least one group of designated channel audios as target data, wherein the designated channel audios include at least two channel audios associated with each other in position.

11. An electronic device, characterized in that: including one or more processors and memory; One or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, wherein when the program code is executed by a processor, the method according to any one of claims 1 to 9 is executed.

Images