Audio acquisition method and device based on artificial intelligence, and unmanned aerial vehicle

Abstract

The invention provides an audio acquisition method and device based on artificial intelligence, and an unmanned aerial vehicle. The method comprises the following steps: obtaining a target audio signal in advance; acquiring a plurality of original audio signals at a plurality of positions; performing separation processing on the plurality of original audio signals to form clean audio signals; comparing the clean audio signal with the target audio signal, and screening out a clean audio signal matched with the target audio signal to form an audio signal to be collected; and transmitting the audio signal to be acquired to a user. Finally, the device and the unmanned aerial vehicle adopting the method separate the noise and the audio signal in the audio information acquisition process, so that the definition of the audio information acquired by the device and the unmanned aerial vehicle is greatly improved, and the audio information acquisition range is also expanded.

Description

[Technical field]

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and in particular to an audio collection method, device and unmanned aerial vehicle based on artificial intelligence. [Background technology]

[0002] With the development of science and technology and low-altitude economy, drones are more widely used in various industries. Drones are gradually applied to logistics, transportation and other industries from traditional limited scenes such as test aircraft and reconnaissance aircraft. In particular, drones are used in police and civilian monitoring, film and television shooting, exploration and exploration tasks. At this time, drones need to be loaded with specific audio collection devices to collect audio from the surrounding environment, and then store the audio or feed it back to the ground control personnel.

[0003] However, the extremely strong self-noise of the drone and the wind noise will interfere with the audio data of the target object. The noise is generated by the rotating motor, circuit, high-speed airflow or surrounding noise when the aircraft is working. As a result, the ratio between signal power and noise power is extremely low in the process of audio collection of existing drones, making it difficult to process the signal; at the same time, the collection range of the audio collection device equipped with the existing drone is too small to collect audio signals at a long distance. [Summary of the invention]

[0004] In view of the above technical problems in the prior art, the present invention provides an audio acquisition method based on artificial intelligence, comprising:

[0005] Acquire a target audio signal in advance;

[0006] Collect multiple original audio signals at multiple locations;

[0007] Separating and processing the plurality of original audio signals to form clean audio signals;

[0008] Comparing the clean audio signal with the target audio signal, and screening out the clean audio signal matching the target audio signal to form a to-be-collected audio signal;

[0009] The audio signal to be collected is transmitted to the user.

[0010] Specifically, the step of collecting multiple original audio signals at multiple locations includes:

[0011] At a first position, collecting a first original audio signal around the first position;

[0012] At a second position, collecting a second original audio signal around the second position;

[0013] The proportions of the noise audio signal contained in the first original audio signal and the second original audio signal are different.

[0014] Specifically, the step of separating and processing the plurality of original audio signals to form a clean audio signal includes:

[0015] independently transmitting the plurality of original audio signals;

[0016] Then, the plurality of original audio signals are aggregated into audio signals of multiple channels;

[0017] The audio signals of the multiple channels are separated and processed to remove the noise audio signal and retain the clean audio signal.

[0018] Specifically, the step of separating and processing the audio signals of the multiple channels to remove the noise audio signal and retain the clean audio signal includes:

[0019] The audio signals of the multiple channels are split into a noise audio signal and a clean audio signal; the audio signals of the multiple channels are defined as follows:

[0020]

[0021] Wherein, Y(k, n) is defined as the audio signal of the multiple channels, H l (k,n) is defined as an acoustic transfer function, X(k,nl) is defined as the noise audio signal, S(k,n) is defined as the clean audio signal; k represents the frequency band of the audio signals of the multiple channels, n represents the sequence number of the short-time frame, and L is the total number of short-time frames;

[0022] Performing vector conversion on the audio signals of the multiple channels, and representing the audio signals of the multiple channels in a vector manner:

[0023] y(k,n)=H(k,n)s(k,n);

[0024] s(k,n)=[S(k,n),X(k,n),···,X(k,n-L+1)] T ;

[0025]

[0026] Wherein, y(k,n) is defined as the audio signals of the multiple channels after vector transformation; T (k,n) is defined as the acoustic transfer function after vector transformation; s(k,n) is defined as the clean audio signal after vector transformation; I means a diagonal matrix with unique values.

[0027] Specifically, the step of comparing the clean audio signal with the target audio signal and screening out the clean audio signal matching the target audio signal to form the audio signal to be collected includes:

[0028] Acquire the clean audio signal;

[0029] Performing correction calculation processing on the clean audio signal, screening out the clean audio signal matching the target audio signal, and defining the clean audio signal matching the target audio signal as a proposed acquisition signal;

[0030] Outputting the signal to be collected;

[0031] The process of performing correction calculation processing on the clean audio signal is as follows:

[0032] e(k,n)=W(k,n)y(k,n);

[0033]

[0034] w(k,n)=V -1 (k,n)e1;

[0035]

[0036] Define the e(k,n) as the information to be collected, W(k,n) as the demixing algorithm, w(k,n) as the vector expression of the demixing algorithm, w T (k,n) is the inverse matrix form of the vector expression of the demixing algorithm, V(k,n) is the weighted covariance matrix, v(b,n) is the activation weight of the value in the nth frame, t(k,b) is the value of the bth basis vector at frequency band k, r(k,n) is the signal variance estimate, α is a fixed constant ranging from 0 to 1; b represents the basis vector, and k represents the frequency band.

[0037] Preferably, the value of the b-th basis vector at the frequency band k and the activation weight of the value at the n-th frame are calculated by an iterative calculation method:

[0038]

[0039] Define the e(k,n) as the information to be collected, v(b,n) as the activation weight of the value in the nth frame, t(k,b) as the value of the bth basis vector at frequency band k, r(k,n) as the signal variance estimate, α as a constant ranging from 0 to 1; b represents the basis vector, k represents the frequency band, and n represents the sequence number of the short-time frame.

[0040] The present invention also provides an audio acquisition device based on artificial intelligence, comprising:

[0041] An acquisition module, used for pre-acquiring a target audio signal;

[0042] A collection module, used for collecting multiple original audio signals at multiple locations;

[0043] A separation module, used for separating and processing the plurality of original audio signals to form clean audio signals;

[0044] A screening module, used for comparing the clean audio signal with the target audio signal, screening out the clean audio signal matching the target audio signal, so as to form a to-be-collected audio signal;

[0045] The transmission module is used to transmit the audio signal to be collected to the user.

[0046] Specifically, the acquisition module includes:

[0047] A first collecting unit, configured to collect, at a first position, a first original audio signal around the first position;

[0048] A second collecting unit is used to collect a second original audio signal around the second position at a second position;

[0049] The proportions of the noise audio signal contained in the first original audio signal and the second original audio signal are different.

[0050] Specifically, the separation module includes:

[0051] An independent transmission unit, used for independently transmitting a plurality of the original audio signals;

[0052] A summarizing unit, used for summarizing the multiple original audio signals into audio signals of multiple channels;

[0053] The separation processing unit is used to separate and process the audio signals of the multiple channels to remove the noise audio signal and retain the clean audio signal.

[0054] Specifically, the separation processing unit includes:

[0055] A splitting component, used for splitting the audio signals of the plurality of channels into a noise audio signal and a clean audio signal;

[0056] The audio signals of the multiple channels are defined as follows:

[0057]

[0058] Wherein, Y(k, n) is defined as the audio signal of the multiple channels, H l(k,n) is defined as an acoustic transfer function, X(k,nl) is defined as the noise audio signal, S(k,n) is defined as the clean audio signal; k represents the frequency band of the audio signals of the multiple channels, n represents the sequence number of the short-time frame, and L is the total number of short-time frames;

[0059] The vector component is used to perform vector conversion on the audio signals of the multiple channels, and to represent the audio signals of the multiple channels in a vector manner:

[0060] y(k,n)=H(k,n)s(k,n);

[0061] s(k,n)=[S(k,n),X(k,n),···,X(k,n-L+1)] T ;

[0062]

[0063] Wherein, y(k,n) is defined as the audio signals of the multiple channels after vector transformation; T (k,n) is defined as the acoustic transfer function after vector transformation; s(k,n) is defined as the clean audio signal after vector transformation; I means a diagonal matrix with unique values;

[0064] Specifically, the screening module includes:

[0065] A cleaning unit, used for acquiring the clean audio signal;

[0066] a correction calculation unit, configured to perform correction calculation processing on the clean audio signal, screen out the clean audio signal matching the target audio signal, and define the clean audio signal matching the target audio signal as a proposed acquisition signal;

[0067] An output unit, used for outputting the signal to be collected;

[0068] The process of performing correction calculation processing on the clean audio signal is as follows:

[0069] e(k,n)=W(k,n)y(k,n);

[0070]

[0071] w(k,n)=V -1 (k,n)e1;

[0072]

[0073] Define the e(k,n) as the information to be collected, W(k,n) as the demixing algorithm, w(k,n) as the vector expression of the demixing algorithm, w T (k,n) is the inverse matrix form of the vector expression of the demixing algorithm, V(k,n) is the weighted covariance matrix, v(b,n) is the activation weight of the value in the nth frame, t(k,b) is the value of the bth basis vector at frequency band k, r(k,n) is the signal variance estimate, α is a fixed constant ranging from 0 to 1; b represents the basis vector, and k represents the frequency band.

[0074] Preferably, the value of the b-th basis vector at the frequency band k and the activation weight of the value at the n-th frame are calculated by an iterative calculation method:

[0075]

[0076] Define the e(k,n) as the information to be collected, v(b,n) as the activation weight of the value in the nth frame, t(k,b) as the value of the bth basis vector at frequency band k, r(k,n) as the signal variance estimate, α as a constant ranging from 0 to 1; b represents the basis vector, k represents the frequency band, and n represents the sequence number of the short-time frame.

[0077] The present invention also provides a drone using an audio collection device based on artificial intelligence, comprising:

[0078] The audio acquisition device based on artificial intelligence;

[0079] body;

[0080] A ground controller, wherein the ground controller is connected to the fuselage wirelessly or by wire;

[0081] The audio collection device is installed on the fuselage and / or the ground controller.

[0082] The present invention provides an audio collection method based on artificial intelligence, comprising the following steps: obtaining a target audio signal in advance; collecting multiple original audio signals at multiple locations; separating and processing the multiple original audio signals to form a clean audio signal; comparing the clean audio signal with the target audio signal, screening out the clean audio signal that matches the target audio signal to form a to-be-collected audio signal; transmitting the to-be-collected audio signal to a user. Ultimately, the device and drone using the method can separate and process the noise and the audio signal during the process of collecting audio information, which greatly improves the clarity of the audio information collected by the device and the drone, and also expands the scope of collecting the audio information.

Brief Description of the Drawings

[0083] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0084] Figure 1 It is a flowchart of an audio acquisition method based on artificial intelligence disclosed by the present invention;

[0085] Figure 2 Yes Figure 1 A schematic diagram of the process of step S2;

[0086] Figure 3 is a flow chart of an embodiment of step S2 of the present invention;

[0087] Figure 4 A schematic flow chart of a preferred embodiment of step S2 of the present invention;

[0088] Figure 5 Yes Figure 1 The schematic flow chart of step S3 shown in FIG.

[0089] Figure 6 Yes Figure 5 The flowchart of step S33 is shown in FIG.

[0090] Figure 7 Yes Figure 1 The schematic flow chart of step S4 shown;

[0091] Figure 8 The present invention also provides a structural schematic diagram of an audio acquisition device 0 based on artificial intelligence;

[0092] Fig. 9 Yes Figure 8 A schematic diagram of the structure of the acquisition module 02;

[0093] Fig.10 Yes Fig. 9 A schematic structural diagram of the separation module 03 of another embodiment;

[0094] Fig.11 Yes Fig.10 The separation processing unit 033;

[0095] Fig.12 Yes Figure 8 The screening module 04;

[0096] Fig.13The present invention provides a drone 09 using an audio collection device based on artificial intelligence. [Specific implementation method]

[0097] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0098] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0099] In addition, in the present invention, descriptions such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0100] In the present invention, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0101] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0102] See also Figure 1 , Figure 1 The present invention discloses an audio acquisition method based on artificial intelligence, which includes the following steps:

[0103] S1, pre-acquire the target audio signal;

[0104] S2, collecting multiple original audio signals at multiple locations;

[0105] S3, separating and processing the plurality of original audio signals to form clean audio signals;

[0106] S4, comparing the clean audio signal with the target audio signal, and screening out the clean audio signal matching the target audio signal to form a to-be-collected audio signal;

[0107] S5: transmitting the audio signal to be collected to the user.

[0108] It should be noted that the present invention collects different original audio signals at different positions, and the proportion of noise contained in different original audio signals is different. One path of the original audio signal uses the noise as a reference, and the other path of the original audio signal uses the intended collected audio signal as a reference. Finally, the two are compared and analyzed to obtain a clear intended collected audio signal.

[0109] See also Figure 2 , Figure 2 is a flow chart of step S2, wherein step S2 comprises:

[0110] S21. At a first position, collecting a first original audio signal around the first position;

[0111] S22, at a second position, collecting a second original audio signal around the second position;

[0112] The proportions of the noise audio signal contained in the first original audio signal and the second original audio signal are different.

[0113] Please note that Figure 3 , Figure 3 is a flow chart of step S2 in another embodiment of the present invention, wherein step S2 comprises:

[0114] In another embodiment of step S2 of the present invention, step S2 includes:

[0115] S021. At a first position, collect a first original audio signal around the first position;

[0116] S022. At a second position, collecting a second original audio signal around the second position;

[0117] S023. At a third position, collecting a third original audio signal around the third position;

[0118] The proportions of noise audio signals included in the first original audio signal, the second original audio signal and the third original audio signal are different.

[0119] It should be noted that the first position, the second position, and the third position represent different parts of the drone.

[0120] Specifically, the first position, the second position and the third position may be any one of the rotating motor of the drone, the tail end of the drone wing, the landing gear of the drone, the front of the drone fuselage, the lower part of the drone fuselage, the rear part of the drone fuselage, etc. Those skilled in the art may set the first position, the second position and the third position according to actual needs, and will not be described in detail here.

[0121] It should be noted that the first original audio signal, the second original audio signal, and the third original audio signal respectively represent any one of the original audio signal collected at the rotating motor of the drone, the original audio signal collected at the tail end of the drone wing, the original audio signal collected at the landing gear of the drone, the original audio signal collected at the front of the drone fuselage, the original audio signal collected at the bottom of the drone fuselage, and the audio signal collected at the rear of the drone fuselage. Those skilled in the art can collect audio signals according to actual needs, which will not be repeated here.

[0122] Please note that Figure 4 , Figure 4 It is a flow chart of step S2 in a preferred embodiment of the present invention, which comprises the following steps:

[0123] S0021. At the rotating motor of the drone, collecting original audio signals around the rotating motor of the drone;

[0124] S0022. At the tail wing of the drone, collecting original audio signals around the tail wing of the drone;

[0125] S0023. At the lower part of the drone body, collect original audio signals around the lower part of the drone body;

[0126] Among them, the original audio signal around the rotating motor of the drone is relatively close to the motor, so it contains more noise generated by the rotation of the motor; the tail of the drone and the lower part of the drone fuselage are far away from the motor, and the noise content of the corresponding original audio collected is different, but the tail of the drone and the lower part of the drone fuselage are located in different positions, and the noise content of the original audio collected by the two is also different. In summary, the proportion of noise of the original audio at the rotating motor of the drone, the tail of the drone, and the lower part of the drone fuselage is different.

[0127] It should be noted that, in this process, due to the different proportions of noise in the original audio at the rotating motor of the drone, the tail wing of the drone and the lower part of the drone fuselage, the different original audios are differentiated and processed. For areas containing more noise, the noise signal is used as a reference to separate the original audio signal; for areas containing more audio to be collected, the audio to be collected is used as a reference to separate the original audio signal.

[0128] It should be noted that the device used to collect audio in the present invention can be a microphone, a detector or other sound receiving device.

[0129] It should be noted that the number of the microphones is greater than or equal to 2.

[0130] In a preferred embodiment of step S2 of the present invention, a low-sensitivity microphone is used at the rotating motor of the drone. Since there is more noise here, a low-sensitivity microphone is used to collect the noise of the rotating motor and high-speed wind noise. A high-sensitivity microphone is used at the lower part of the drone fuselage or the tail of the drone. Since there is less noise here, a high-sensitivity microphone is used to collect the target audio signal in the original audio signal.

[0131] It should be noted that the target audio signal may be a specific audio signal that needs to be collected, such as a human voice, a certain animal call, a certain natural sound, etc., which are not listed here one by one, and those skilled in the art may freely choose according to actual needs.

[0132] See also Figure 5 , Figure 5 Yes Figure 1 The schematic flow chart of step S3 shown in FIG. 1 comprises the following steps:

[0133] S31, independently transmitting the plurality of original audio signals;

[0134] S32, then aggregating the multiple original audio signals into audio signals of multiple channels;

[0135] S33: Separate and process the audio signals of the multiple channels to remove the noise audio signal and retain the clean audio signal.

[0136] See also Figure 6 , Figure 6 Yes Figure 5 The flowchart of step S33 shown in FIG. 1 includes the following steps:

[0137] S331, splitting the audio signals of the multiple channels into a noise audio signal and a clean audio signal; the audio signals of the multiple channels are defined as follows:

[0138]

[0139] Wherein, Y(k, n) is defined as the audio signal of the multiple channels, H l (k,n) is defined as an acoustic transfer function, X(k,nl) is defined as the noise audio signal, S(k,n) is defined as the clean audio signal; k represents the frequency band of the audio signals of the multiple channels, n represents the sequence number of the short-time frame, and L is the total number of short-time frames;

[0140] S332, performing vector conversion on the audio signals of the multiple channels; converting the audio signals representing the multiple channels in a vector manner:

[0141] y(k,n)=H(k,n)s(k,n);

[0142] s(k,n)=[S(k,n),X(k,n),···,X(k,n-L+1)] T ;

[0143]

[0144] Wherein, y(k,n) is defined as the audio signals of the multiple channels after vector transformation; T (k,n) is defined as the acoustic transfer function after vector transformation; s(k,n) is defined as the clean audio signal after vector transformation; I means a diagonal matrix with unique values.

[0145] It should be noted that the present invention adopts a modeling method to model the chaotic original audio signal into a clean audio signal and a noisy audio signal, and then separates the clean audio signal from the noisy audio signal.

[0146] See also Figure 7 , Figure 7 Yes Figure 1 The schematic flow chart of step S4 shown in FIG. 1 comprises the following steps:

[0147] S41, obtaining the clean audio signal;

[0148] S42, performing correction calculation processing on the clean audio signal, screening out the clean audio signal matching the target audio signal, and defining the clean audio signal matching the target audio signal as a proposed acquisition signal;

[0149] S43, outputting the proposed acquisition signal;

[0150] The process of performing correction calculation processing on the clean audio signal is as follows:

[0151] e(k,n)=W(k,n)y(k,n);

[0152]

[0153] w(k,n)=V -1 (k,n)e1;

[0154]

[0155] Define the e(k,n) as the information to be collected, W(k,n) as the demixing algorithm, w(k,n) as the vector expression of the demixing algorithm, w T (k,n) is the inverse matrix form of the vector expression of the demixing algorithm, V(k,n) is the weighted covariance matrix, v(b,n) is the activation weight of the value in the nth frame, t(k,b) is the value of the bth basis vector at frequency band k, r(k,n) is the signal variance estimate, α is a fixed constant ranging from 0 to 1; b represents the basis vector, and k represents the frequency band.

[0156] In a preferred embodiment, the value of the b-th basis vector at the frequency band k and the activation weight of the value at the n-th frame are calculated by an iterative calculation method:

[0157]

[0158] Define the e(k,n) as the information to be collected, v(b,n) as the activation weight of the value in the nth frame, t(k,b) as the value of the bth basis vector at frequency band k, r(k,n) as the signal variance estimate, α as a constant ranging from 0 to 1; b represents the basis vector, k represents the frequency band, and n represents the sequence number of the short-time frame.

[0159] See also Figure 8 , Figure 8 The present invention also provides a structural schematic diagram of an audio collection device 0 based on artificial intelligence, and the audio collection device 0 includes:

[0160] Acquisition module 01, used to pre-acquire the target audio signal;

[0161] The acquisition module 02 is used to acquire multiple original audio signals at multiple locations;

[0162] A separation module 03, used for separating and processing the plurality of original audio signals to form clean audio signals;

[0163] A screening module 04 is used to compare the clean audio signal with the target audio signal, and screen out the clean audio signal that matches the target audio signal to form a to-be-collected audio signal;

[0164] The transmission module 05 is used to transmit the audio signal to be collected to the user.

[0165] See also Fig. 9 , Fig. 9 Yes Figure 8 The structural diagram of the acquisition module 02, the acquisition module 02 comprises:

[0166] A first collecting unit 021 is used to collect a first original audio signal around a first position at a first position;

[0167] A second collecting unit 022 is used to collect a second original audio signal around the second position at a second position;

[0168] A second collecting unit 023 is used to collect a third original audio signal around the third position at a third position;

[0169] The proportions of noise audio signals included in the first original audio signal, the second original audio signal and the third original audio signal are different.

[0170] See also Fig.10 , Fig.10 Yes Fig. 9 A schematic diagram of the structure of the separation module 03 of another embodiment. The separation module 03 comprises:

[0171] An independent transmission unit 031, used for independently transmitting a plurality of the original audio signals;

[0172] A summarizing unit 032, configured to aggregate the plurality of original audio signals into audio signals of a plurality of channels;

[0173] The separation processing unit 033 is used to separate the audio signals of the multiple channels to remove the noise audio signal and retain the clean audio signal.

[0174] See also Fig.11 , Fig.11 Yes Fig.10 The separation processing unit 033 includes:

[0175] A splitting component 0331, used for splitting the audio signals of the multiple channels into a noise audio signal and a clean audio signal;

[0176] The audio signals of the multiple channels are defined as follows:

[0177]

[0178] Wherein, Y(k, n) is defined as the audio signal of the multiple channels, H l (k,n) is defined as an acoustic transfer function, X(k,nl) is defined as the noise audio signal, S(k,n) is defined as the clean audio signal; k represents the frequency band of the audio signals of the multiple channels, n represents the sequence number of the short-time frame, and L is the total number of short-time frames;

[0179] The vector component 0332 is used to perform vector conversion on the audio signals of the multiple channels, and to represent the audio signals of the multiple channels in a vector manner:

[0180] y(k,n)=H(k,n)s(k,n);

[0181] s(k,n)=[S(k,n),X(k,n),···,X(k,n-L+1)] T ;

[0182]

[0183] Wherein, y(k,n) is defined as the audio signals of the multiple channels after vector transformation; T (k,n) is defined as the acoustic transfer function after vector transformation; s(k,n) is defined as the clean audio signal after vector transformation; I means a diagonal matrix with unique values;

[0184] See also Fig.12 , Fig.12 Yes Figure 8 The screening module 04 comprises:

[0185] A cleaning unit 041, used to obtain the clean audio signal;

[0186] A correction calculation unit 042 is used to perform correction calculation processing on the clean audio signal, filter out the clean audio signal matching the target audio signal, and define the clean audio signal matching the target audio signal as a proposed acquisition signal;

[0187] An output unit 043, used for outputting the proposed acquisition signal;

[0188] The process of performing correction calculation processing on the clean audio signal is as follows:

[0189] e(k,n)=W(k,n)y(k,n);

[0190]

[0191] w(k,n)=V -1 (k,n)e1;

[0192]

[0193] Define the e(k,n) as the information to be collected, W(k,n) as the demixing algorithm, w(k,n) as the vector expression of the demixing algorithm, w T (k,n) is the inverse matrix form of the vector expression of the demixing algorithm, V(k,n) is the weighted covariance matrix, v(b,n) is the activation weight of the value in the nth frame, t(k,b) is the value of the bth basis vector at frequency band k, r(k,n) is the signal variance estimate, α is a fixed constant ranging from 0 to 1; b represents the basis vector, and k represents the frequency band.

[0194] Preferably, the value of the b-th basis vector at the frequency band k and the activation weight of the value at the n-th frame are calculated by an iterative calculation method:

[0195]

[0196]

[0197] Define the e(k,n) as the information to be collected, v(b,n) as the activation weight of the value in the nth frame, t(k,b) as the value of the bth basis vector at frequency band k, r(k,n) as the signal variance estimate, α as a constant ranging from 0 to 1; b represents the basis vector, k represents the frequency band, and n represents the sequence number of the short-time frame.

[0198] See also Fig.13 , Fig.13 The present invention provides a drone 09 using an audio collection device based on artificial intelligence, and the drone 09 includes:

[0199] The audio collection device 08 based on artificial intelligence includes: an acquisition module for pre-acquiring a target audio signal; a collection module for collecting multiple original audio signals at multiple locations; a separation module for separating and processing the multiple original audio signals to form a clean audio signal; a screening module for comparing the clean audio signal with the target audio signal, screening out the clean audio signal matching the target audio signal to form a to-be-collected audio signal; and a transmission module for transmitting the to-be-collected audio signal to a user. The fuselage 07; the ground controller 06, the ground controller 06 and the fuselage 07 are connected wirelessly or by wire; the audio collection device 08 based on artificial intelligence is installed on the fuselage 07 and / or the ground controller 06.

[0200] In one embodiment of the present invention, the artificial intelligence-based audio acquisition device 08 is installed on the body 07 .

[0201] In another embodiment of the present invention, the artificial intelligence-based audio acquisition device 08 is installed on the ground controller 06 .

[0202] In another embodiment of the present invention, the artificial intelligence-based audio acquisition device 08 is installed on the fuselage 07 and the ground controller 06, that is, a part of the modules, units or components of the artificial intelligence-based audio acquisition device 08 are installed in the fuselage 07, and a part of them are installed in the ground controller 06.

[0203] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An audio acquisition method based on artificial intelligence, characterized in that: include: Acquire a target audio signal in advance; Collect multiple original audio signals at multiple locations; Separating and processing the plurality of original audio signals to form clean audio signals; Comparing the clean audio signal with the target audio signal, and screening out the clean audio signal matching the target audio signal to form a to-be-collected audio signal; The audio signal to be collected is transmitted to the user.

2. The audio acquisition method based on artificial intelligence according to claim 1, characterized in that: The step of collecting a plurality of original audio signals at a plurality of locations comprises: At a first position, collecting a first original audio signal around the first position; At a second position, collecting a second original audio signal around the second position; The proportions of the noise audio signal contained in the first original audio signal and the second original audio signal are different.

3. The audio acquisition method based on artificial intelligence according to claim 1, characterized in that: The step of separating and processing the plurality of original audio signals to form a clean audio signal comprises: independently transmitting the plurality of original audio signals; Then, the plurality of original audio signals are aggregated into audio signals of multiple channels; The audio signals of the multiple channels are separated and processed to remove the noise audio signal and retain the clean audio signal.

4. The audio acquisition method based on artificial intelligence as claimed in claim 3, characterized in that: The step of separating and processing the audio signals of the multiple channels to remove the noise audio signal and retain the clean audio signal comprises: The audio signals of the multiple channels are split into a noise audio signal and a clean audio signal; the audio signals of the multiple channels are defined as follows: Wherein, Y(k, n) is defined as the audio signal of the multiple channels, H l (k,n) is defined as an acoustic transfer function, X(k,nl) is defined as the noise audio signal, S(k,n) is defined as the clean audio signal; k represents the frequency band of the audio signals of the multiple channels, n represents the sequence number of the short-time frame, and L is the total number of short-time frames; Performing vector conversion on the audio signals of the multiple channels, and representing the audio signals of the multiple channels in a vector manner: y(k,n)=H(k,n)s(k,n); s(k,n)=[S(k,n),X(k,n),···,X(k,n-L+1)] T ; Wherein, y(k,n) is defined as the audio signals of the multiple channels after vector transformation; T (k,n) is defined as the acoustic transfer function after vector transformation; s(k,n) is defined as the clean audio signal after vector transformation; I means a diagonal matrix with unique values.

5. The audio acquisition method based on artificial intelligence as claimed in claim 4, characterized in that: The step of comparing the clean audio signal with the target audio signal and selecting the clean audio signal matching the target audio signal to form the audio signal to be collected includes: Acquire the clean audio signal; Performing correction calculation processing on the clean audio signal, screening out the clean audio signal matching the target audio signal, and defining the clean audio signal matching the target audio signal as a proposed acquisition signal; Outputting the signal to be collected; The process of performing correction calculation processing on the clean audio signal is as follows: e(k,n)=W(k,n)y(k,n); in(k,n)=V -1 (k,n)e1; Define ek(,n) as the information to be collected, W(k,n) as the demixing algorithm, w(k,n) as the vector expression of the demixing algorithm, w T (k,n) is the inverse matrix form of the vector expression of the demixing algorithm, V(k,n) is the weighted covariance matrix, v(b,n) is the activation weight of the value in the nth frame, t(k,b) is the value of the bth basis vector at frequency band k, r(k,n) is the signal variance estimate, α is a fixed constant ranging from 0 to 1; b represents the basis vector, and k represents the frequency band.

6. An audio acquisition device based on artificial intelligence, characterized in that: include: An acquisition module, used for pre-acquiring a target audio signal; A collection module, used for collecting multiple original audio signals at multiple locations; A separation module, used for separating and processing the plurality of original audio signals to form clean audio signals; A screening module, used for comparing the clean audio signal with the target audio signal, screening out the clean audio signal matching the target audio signal, so as to form a to-be-collected audio signal; The transmission module is used to transmit the audio signal to be collected to the user.

7. The audio acquisition device based on artificial intelligence as claimed in claim 6, characterized in that: The acquisition module comprises: A first collecting unit, configured to collect, at a first position, a first original audio signal around the first position; A second collecting unit is used to collect a second original audio signal around the second position at a second position; The proportions of the noise audio signal contained in the first original audio signal and the second original audio signal are different.

8. The audio acquisition device based on artificial intelligence as claimed in claim 6, characterized in that: The separation module comprises: An independent transmission unit, used for independently transmitting a plurality of the original audio signals; A summarizing unit, used for summarizing the multiple original audio signals into audio signals of multiple channels; The separation processing unit is used to separate and process the audio signals of the multiple channels to remove the noise audio signal and retain the clean audio signal. The separation processing unit comprises: A splitting component, used for splitting the audio signals of the plurality of channels into a noise audio signal and a clean audio signal; The audio signals of the multiple channels are defined as follows: Wherein, Y(k, n) is defined as the audio signal of the multiple channels, H l (k,n) is defined as an acoustic transfer function, X(k,nl) is defined as the noise audio signal, S(k,n) is defined as the clean audio signal; k represents the frequency band of the audio signals of the multiple channels, n represents the sequence number of the short-time frame, and L is the total number of short-time frames; The vector component is used to perform vector conversion on the audio signals of the multiple channels, and to represent the audio signals of the multiple channels in a vector manner: y(k,n)=H(k,n)s(k,n); s(k,n)=[S(k,n),X(k,n),···,X(k,n-L+1)] T ; Wherein, y(k,n) is defined as the audio signals of the multiple channels after vector transformation; T (k,n) is defined as the acoustic transfer function after vector transformation; s(k,n) is defined as the clean audio signal after vector transformation; I means a diagonal matrix with unique values.

9. The audio acquisition device based on artificial intelligence as claimed in claim 8, characterized in that: The screening module comprises: A cleaning unit, used for acquiring the clean audio signal; a correction calculation unit, configured to perform correction calculation processing on the clean audio signal, screen out the clean audio signal matching the target audio signal, and define the clean audio signal matching the target audio signal as a proposed acquisition signal; An output unit, used for outputting the signal to be collected; The process of performing correction calculation processing on the clean audio signal is as follows: e(k,n)=W(k,n)y(k,n); in(k,n)=V -1 (k,n)e1; Define the e(k,n) as the information to be collected, W(k,n) as the demixing algorithm, w(k,n) as the vector expression of the demixing algorithm, w T (k,n) is the inverse matrix form of the vector expression of the demixing algorithm, V(k,n) is the weighted covariance matrix, v(b,n) is the activation weight of the value in the nth frame, t(k,b) is the value of the bth basis vector at frequency band k, r(k,n) is the signal variance estimate, α is a fixed constant ranging from 0 to 1; b represents the basis vector, and k represents the frequency band.

10. A drone using an audio acquisition device based on artificial intelligence, characterized in that: include: The audio acquisition device based on artificial intelligence as claimed in claim 6; body; A ground controller, wherein the ground controller is connected to the fuselage wirelessly or by wire; The audio collection device is installed on the fuselage and / or the ground controller.

Images