A far-field noise event evidence collection system and method for drone-mounted sound pickup equipment

Abstract

The present invention discloses a far-field noise event forensics system and method for a drone-mounted sound pickup device. The system includes a multi-rotor drone, a background microphone, a far-field noise pickup module, a fishing module, a vibration reduction module, and an audio processing module. The background microphone collects background noise. The multi-rotor drone is provided with a video acquisition module. The fishing module adjusts the fishing height of the far-field noise pickup module. The far-field noise pickup module picks up far-field noise at different altitudes. The audio processing module performs spectrum analysis on the background noise based on short-time Fourier transform, performs Fourier transform on the background noise, and then performs iterative simulation in combination with a priori tuned active noise reduction algorithm to generate an inverted signal of the background noise signal, which is superimposed with the far-field noise signal to obtain effective target noise audio, and converted into sound pressure level data. The present invention performs multi-source audio and video forensics on far-field noise events based on passive noise reduction and active noise reduction, which can more efficiently achieve noise reduction effects and improve signal-to-noise ratio.

Description

Technical Field

[0001] The present invention relates to the technical field of noise data processing, and in particular to a far-field noise event evidence collection system and method for an unmanned aerial vehicle-mounted sound pickup device. Background Art

[0002] The flight missions of existing drones are usually based on multi-dimensional camera functions, and very few drones have recording functions. The recording objects of drones with recording functions are mostly daily sound sources, which have low sound pressure levels and are difficult to pick up. In addition, drones with recording functions do not consider the connection method between the audio recording module and the drone, as well as the vibration of the drone and the interference to the audio recording. The vibration will affect the stability of the sound quality of the audio recording. Drones with recording functions rarely use a composite noise reduction method of active noise reduction and passive noise reduction.

[0003] In the field of aircraft equipped with microphones, the flight altitude of the aircraft is much higher than 30m, so the noise of the drone equipped with microphones for evidence collection is in the far-field sound field. At this time, the microphone's ability to pick up ground sound sources will be greatly weakened. The impact of far-field ground sound sources on the microphone is smaller than the background noise of the drone. In order to meet the requirements of the effective distance of the microphone, the drone chooses to fly at a lower altitude, which will be limited by the ground reflection of the rotor noise and the increased risk of obstacle avoidance safety.

[0004] Furthermore, drones with recording capabilities mostly use ANC active noise reduction, which requires the installation of multiple secondary sound source generators to counteract the background noise of the aircraft. However, the sound waves of the secondary sound source generators are adaptively adjusted due to different installation positions, and the anti-phase antagonistic noise waves generated by them are difficult to perfectly cancel out the original background sound waves at the focus of the microphone. Instead, due to the phase difference, the sound waves are superimposed to generate greater noise.

[0005] Therefore, in today's urban environmental noise source tracing observation and management system, there is an urgent need to use drone-mounted sound pickup equipment to capture and quantify far-field noise, so as to achieve a highly mobile, vertical observation, and audio and video multi-source evidence collection method for complaints of excessive noise. Summary of the invention

[0006] In order to overcome the defects and shortcomings of the prior art, the present invention provides a far-field noise event forensics system and method for unmanned aerial vehicle-mounted sound pickup equipment, which performs passive noise reduction by arranging noise reduction blades, noise reduction sound collection covers, vibration reduction modules and fishing modules, and generates anti-noise signals by priori tuning of the active noise reduction algorithm more quickly and iteratively, performs anti-noise cancellation after the acoustic-electric signal conversion, and finally obtains effective audio by inverse Fourier transform, which can more effectively improve the signal-to-noise ratio of the signal, and finally realizes multi-source audio and video forensics of far-field noise events based on a composite noise reduction method of passive noise reduction and active noise reduction.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] The present invention provides a far-field noise event evidence collection system of an unmanned aerial vehicle-mounted sound pickup device, comprising: a multi-rotor unmanned aerial vehicle, a background microphone, a far-field noise pickup module, a fishing module, a vibration reduction module and an audio processing module;

[0009] The background microphone is arranged on the top of the multi-rotor drone, and is used to collect the background noise of the multi-rotor drone at different flight altitudes and in different scenarios;

[0010] The multi-rotor drone is provided with a video acquisition module for synchronously acquiring on-site video data, the multi-rotor drone is connected to a fishing module via a vibration reduction module, the fishing module is connected to a far-field noise pickup module, the fishing module is used to adjust the fishing height of the far-field noise pickup module, and the far-field noise pickup module is used to pick up far-field noise at different altitudes;

[0011] The audio processing module includes a spectrum analysis unit, an active noise reduction unit, a data analysis and conversion unit, and a data storage unit;

[0012] The spectrum analysis unit is used to perform spectrum analysis on the background noise based on short-time Fourier transform, and transmit the spectrum analysis result to the active noise reduction unit;

[0013] The active noise reduction unit is used to perform Fourier transform on the background noise and then perform iterative simulation in combination with the prior-tuned active noise reduction algorithm to generate an inverted signal of the background noise signal, achieve phase cancellation by superimposing the time domain and far-field noise signals, obtain effective target noise audio by inverse Fourier transform, and transmit it to the data analysis and conversion unit;

[0014] The data analysis and conversion unit is used to analyze and convert the effective target noise audio, convert the time domain amplitude into sound pressure level data, and transmit it to the data storage unit;

[0015] The data storage unit is used to receive and store the analysis and conversion results of the data analysis and conversion unit, and to store the video data of the ground far-field noise event collected by the video acquisition module.

[0016] As a preferred technical solution, the far-field noise pickup module includes a fixed rod, a sound card, a noise reduction and sound focusing cover and a pickup microphone. One end of the fixed rod is connected to the fishing module, and the other end of the fixed rod is connected to the noise reduction and sound focusing cover. The pickup microphone is arranged at the internal reflection focus of the noise reduction and sound focusing cover to pick up far-field noise. The sound card is used to convert the electrical signals collected by the background microphone and the pickup microphone into digital signals. The noise reduction and sound focusing cover is used to reduce the background noise transmitted above the drone and converge the far-field noise transmitted on the ground.

[0017] As a preferred technical solution, the noise reduction and sound gathering cover is provided with a metal shell, a sound insulation layer, a sound absorbing layer and a reflective layer from the outside to the inside. The sound insulation layer is used to isolate the background noise transmitted from the top of the multi-rotor drone, the sound absorbing layer is used to absorb the remaining background noise passing through the sound insulation layer, and the reflective layer is used to reflect far-field sound waves and gather the far-field sound waves at the focus inside the noise reduction and sound gathering cover structure.

[0018] As a preferred technical solution, the fishing module includes a fishing retractable controller, a brushless motor, a battery, a fishing line reel and a fishing line. The fishing retractable controller is electrically connected to the brushless motor, the output shaft of the brushless motor is connected to the fishing line reel, the battery is electrically connected to the brushless motor, one end of the fishing line is connected to the fishing line reel, and the other end of the fishing line is connected to the far-field noise pickup module. The fishing retractable controller controls the output shaft of the brushless motor to rotate forward or reverse through a retractable control signal, drives the fishing line reel to rotate, and adjusts the fishing height of the far-field noise pickup module through the fishing line.

[0019] As a preferred technical solution, the multi-rotor drone is provided with ultra-carbon fiber noise reduction blades.

[0020] As a preferred technical solution, the background microphone adopts an omnidirectional microphone, the sound pickup microphone adopts a unidirectional microphone, and both the background microphone and the sound pickup microphone are provided with windproof cotton.

[0021] The present invention also provides a method for collecting evidence of far-field noise events of unmanned aerial vehicle sound pickup equipment, which is provided with the above-mentioned far-field noise event collection system of unmanned aerial vehicle sound pickup equipment, and comprises the following steps:

[0022] The flight parameter data of the multi-rotor drone is collected at different altitudes and different fishing heights, the first background noise is obtained based on the background microphone, and a parameter database of different flight monitoring scenarios is constructed, and the fishing height of the fishing module is set according to the sound attenuation algorithm, and the flight altitude of the multi-rotor drone is set;

[0023] Perform flight monitoring tasks, collect the second background source noise signal of the drone during the flight monitoring task based on the background microphone, transmit the second background source noise signal data to the audio processing module, obtain the sound power spectrum through spectrum analysis, and obtain the frequency band of the second background source noise;

[0024] Perform flight monitoring tasks and collect ground far-field noise signals during flight monitoring tasks based on microphones;

[0025] Based on the prior-tuned active noise reduction algorithm, an inverted signal of the second background source noise signal is generated, and superimposed with the ground far-field noise signal to generate an anti-noise signal;

[0026] The anti-noise signal is converted into an audio signal file for storage, and the corresponding simulated sound pressure level is calculated and generated.

[0027] As a preferred technical solution, the sound attenuation algorithm specifically includes:

[0028] When the wavelength is much larger than the size of the sound source and in the far field, the sound source is regarded as a point sound source. Under the far field conditions of the free sound field, the sound power level is used as the evaluation quantity of the point sound source, and the attenuation formula is expressed as:

[0029]

[0030] Among them, ΔL represents the attenuation value caused by the increase in distance, the unit is dB, r represents the distance from the point sound source to the sound receiving point, the unit is m;

[0031] At a distance from a point sound source, the attenuation value from r1 to r2 is expressed as:

[0032] ΔL=20lg(r1 / r2)

[0033] Among them, r1 represents the distance between the sound pickup microphone and the noise reduction hood, and r2 represents the distance from the drone to the sound pickup microphone.

[0034] As a preferred technical solution, the active noise reduction algorithm based on prior tuning generates an inverted signal of the second background source noise signal, and superimposes it with the ground far-field noise signal to generate an anti-noise signal. The specific steps include:

[0035] Using the filter tap weight coefficient at time n-1, the filter tap weight coefficient at time n is obtained through iterative updating;

[0036] The cost function is constructed based on the exponentially weighted sum of squared errors, expressed as:

[0037] J(n)=∑λ n-i |d(i)-w H (n)*x(i)| 2

[0038] Among them, λ represents the forgetting factor, w represents the weight, d(i) represents the ground far-field noise signal sequence picked up by the pickup microphone, and x(i) represents the background noise sequence collected by the background microphone.

[0039] Taking partial derivative of the weights in the cost function, the optimal solution is:

[0040] w(n)=R -1 (n)r(n)

[0041]

[0042] Iterative update, using the matrix inversion lemma:

[0043]

[0044] Among them, k(n) represents the gain vector, and P(n) represents the inverse matrix;

[0045] Update weights:

[0046] w(n)=w(n-1)+k(n)ε * (n)

[0047] Update error:

[0048] ε(n)=d(n)-w H (n-1)x(n)

[0049] Use the updated filter tap coefficients to calculate and obtain the analog signal w H (n-1)x(n), the analog signal inversion signal is superimposed with the ground far-field noise signal sequence to obtain the anti-noise signal ε(n).

[0050] As a preferred technical solution, the calculation to generate the corresponding simulated sound pressure level specifically includes:

[0051]

[0052] Among them, P′ e Indicates the effective sound pressure level, P e It represents the instantaneous sound pressure collected within the set time, x is the signal amplitude converted by the microphone, N is the number of time domain sampling points, P ref Indicates the reference sound pressure, P AOP is the maximum sound pressure level of the hardware microphone;

[0053] The effective sound pressure level P′ e Substituting into the equivalent sound pressure level formula, we get:

[0054]

[0055] Here, SPL stands for sound pressure level.

[0056] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0057] (1) The present invention realizes the evidence collection of target far-field noise based on a composite noise reduction method of passive noise reduction and active noise reduction. Specifically, the present invention performs passive noise reduction by arranging noise reduction blades, noise reduction sound collecting hoods, vibration reduction modules and fishing modules, thereby greatly reducing the background noise and other high-frequency noises transmitted above the multi-rotor UAV. Passive noise reduction absorbs and isolates the background noise transmitted by the UAV above. At the same time, the ground target noise signal is amplified according to the principle of sound reflection, thereby further improving the effective signal-to-noise ratio. Specifically, the noise reduction blade structure reduces the eddy wind noise and vibration generated by the blades during the UAV flight mission. The noise reduction sound collecting hood is provided with a metal shell, a sound insulation layer, a sound absorbing layer and a reflection layer from the outside to the inside. The noise reduction sound collecting hood effectively isolates the rotor noise and machine noise transmitted above the UAV. At the same time, the pickup microphone is arranged in the noise reduction sound collecting hood, which can better capture the effective audio signal of the target on the ground, thereby greatly improving the signal-to-noise ratio.

[0058] The present invention optimizes the active noise reduction algorithm a priori, iterates and generates anti-noise signals more quickly, performs anti-noise cancellation after the acoustic-electrical signal is converted, and finally obtains effective audio through inverse Fourier transform. It can more effectively improve the signal-to-noise ratio of the signal, complete the noise reduction process in the DSP chip, and at the same time abandon the secondary sound source generator in the traditional ANC, avoiding the complexity of the system and reducing the phase deviation caused by the hardware medium in the actual propagation process, and can more efficiently achieve the noise reduction effect.

[0059] (2) The multi-rotor UAV of the present invention is connected to the fishing module through a vibration reduction module to reduce the impact of vibration caused by the flight of the multi-rotor UAV. The fishing module is connected to the far-field noise pickup module, and the fishing height of the far-field noise pickup module can be freely adjusted to allow the far-field noise pickup module to perform fixed-point fishing for evidence collection, thereby effectively increasing the distance between the background sound source of the multi-rotor UAV and the far-field noise pickup module, and reducing the distance between the far-field noise pickup module and the ground noise source. According to the noise attenuation theory, the signal-to-noise ratio of the audio that the pickup can pick up noise can be greatly improved. At the same time, the flight altitude of the multi-rotor UAV is also greatly improved, reducing the rotor noise reflection and obstacle avoidance safety risks of the multi-rotor UAV, and improving the scene adaptability of the multi-rotor UAV to the evidence collection of far-field noise nuisance incidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a structural schematic diagram of a far-field noise event evidence collection system of a drone-mounted sound pickup device of the present invention;

[0061] Figure 2 It is a structural schematic diagram of the fishing module of the present invention;

[0062] Figure 3 It is a structural schematic diagram of the far-field noise pickup module of the present invention;

[0063] Figure 4 It is a schematic diagram of the process of the method for collecting evidence of far-field noise events of the drone-mounted sound pickup device of the present invention;

[0064] Figure 5 Schematic diagram of the sound power spectrum of the background noise of the present invention;

[0065] Figure 6 It is a schematic diagram of the process of a priori tuning of the active noise reduction algorithm of the present invention;

[0066] Among them, 1- multi-rotor drone, 2- background microphone, 3- far-field noise pickup module, 4- fishing module, 5- video acquisition module, 6- ultra-carbon fiber noise reduction blades, 7- vibration reduction module, 8- audio processing module, 9- fishing reel controller, 10- brushless motor, 11- battery, 12- fishing reel, 13- fixed rod, 14- sound card, 15- noise reduction sound collection cover, 16- pickup microphone, 17- metal shell, 18- sound insulation layer, 19- sound absorption layer, 20- reflection layer. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] Example 1

[0069] like Figure 1 As shown, this embodiment provides a far-field noise event forensics system for a drone-mounted sound pickup device, comprising: a multi-rotor drone 1, a background microphone 2, a far-field noise pickup module 3, a fishing module 4, a vibration reduction module 7 and an audio processing module 8;

[0070] Among them, the background microphone 2 is arranged on the top of the multi-rotor drone 1, and the background microphone is used to collect the background noise of the multi-rotor drone at different flight altitudes and in different scenarios; the multi-rotor drone is provided with a video acquisition module 5 for synchronously acquiring on-site video data, the multi-rotor drone is connected to the fishing module 4 through the vibration reduction module 7, and the fishing module 4 is connected to the far-field noise pickup module 3;

[0071] In this embodiment, the multi-rotor UAV is provided with ultra-carbon fiber noise reduction blades 6 to reduce the eddy wind noise and vibration generated by the blades during the UAV flight mission.

[0072] like Figure 2As shown, the fishing module includes: a fishing retractable controller 9, a brushless motor 10, a battery 11, a fishing line reel 12 and a fishing line, wherein the fishing retractable controller 9 is electrically connected to the brushless motor 10, the output shaft of the brushless motor 10 is connected to the fishing line reel 12, the battery 11 is electrically connected to the brushless motor 10, one end of the fishing line is connected to the fishing line reel 12, and the other end of the fishing line is connected to the fixed rod 13 of the far-field noise pickup module; the fishing retractable controller controls the output shaft of the brushless motor to rotate forward or reverse through a retractable control signal, thereby driving the fishing line reel to rotate, and adjusting the fishing height of the far-field noise pickup module through the fishing line. The flight altitude of the multi-rotor UAV and the fishing height of the far-field noise pickup module can be freely adjusted according to the complexity of the application area scene, and far-field noise at different altitudes can be picked up, thereby ensuring the safety, stability and applicability of the UAV.

[0073] like Figure 3 As shown, the far-field noise pickup module includes: a fixed rod 13, a sound card 14, a noise reduction and sound collection cover 15, and a pickup microphone 16, wherein one end of the fixed rod 13 is connected to the fishing module, and the other end of the fixed rod 13 is connected to the noise reduction and sound collection cover 15, and the pickup microphone 16 is arranged in the noise reduction and sound collection cover 15 for picking up far-field noise;

[0074] In this embodiment, the noise reduction and sound gathering cover reduces the background noise transmitted above the multi-rotor drone during the sound picking task, and gathers the target noise transmitted on the ground; the noise reduction and sound gathering cover is provided with: a metal shell 17, a sound insulation layer 18, a sound absorption layer 19, and a reflection layer 20 from the outside to the inside, wherein the metal shell 17 is preferably an aluminum alloy shell, which plays a role in wind and rain protection, the sound insulation layer 18 is preferably a high-efficiency damping sound insulation felt, which is used to isolate the background noise transmitted from the top of the multi-rotor drone, the sound absorption layer 19 is preferably melamine sponge plus melamine foam, and the sound absorption coefficient is 0.9, which is used to absorb the remaining background noise passing through the sound insulation layer, and the reflection layer 20 is preferably aluminum foil, which is used to reflect far-field sound waves and gather the far-field sound waves at the internal focus of the noise reduction and sound gathering cover structure; after structural design, the internal reflection focus of the noise reduction and sound gathering cover is initially determined, and after 3D printing, the internal reflection focus is verified by emitting parallel light;

[0075] In this embodiment, the sound card is used to convert the electrical signals collected by the background microphone and the pickup microphone into digital signals through analog-to-digital conversion;

[0076] In this embodiment, the vibration reduction module uses multiple vibration reduction balls to reduce the impact of drone vibration on the far-field noise pickup module acquiring audio signals;

[0077] In this embodiment, the background microphone is preferably an omnidirectional microphone, which can better pick up the background noise of the near-field drone at different altitudes, and is equipped with windproof cotton to filter the wind sound;

[0078] In this embodiment, the pickup microphone is a unidirectional microphone and is provided with a windproof cotton for filtering wind sound. The pickup microphone is arranged at the internal reflection focus of the noise reduction and focusing cover to pick up far-field noise.

[0079] In this embodiment, the audio processing module 8 includes a spectrum analysis unit, an active noise reduction unit, a data analysis and conversion unit, and a data storage unit;

[0080] The spectrum analysis unit performs spectrum analysis on the audio of the picked-up background noise based on short-time Fourier transform, and transmits the analysis result to the active noise reduction unit;

[0081] The active noise reduction unit performs Fourier transform on the background noise collected by the background microphone and then performs iterative simulation based on the prior-tuned active noise reduction algorithm to generate an inverted signal of the background noise signal. Finally, in the DSP chip, the phase is canceled by superimposing the time domain with the ground far-field noise signal. Then, the effective target noise audio for low-frequency noise reduction is obtained through inverse Fourier transform and transmitted to the data analysis and conversion unit.

[0082] The data analysis and conversion unit analyzes and converts the effective target noise audio, converts the time domain amplitude into sound pressure level data, which can better reflect the change trend of noise pollution in the mission area after the drone conducts multi-point far-field evidence collection and draws the area, and finally transmits the data analysis results and audio to the data storage unit;

[0083] The data storage unit receives the audio and video evidence and data analysis results of the noise event from the video acquisition module and the data analysis conversion unit. Specifically, the video acquisition module (such as a camera) carried by the drone can shoot, record, store and transmit the video and audio evidence data of the noise event in the flight monitoring mission, and finally transmit it to the data storage unit;

[0084] In this embodiment, the audio file after noise reduction is the sound collected by the digital microphone. Due to the different sensitivities of different microphones, the time domain sampling point values ​​collected by the microphone are converted to RMS values, and then the maximum sound pressure level of the microphone is added to obtain the sound pressure level data of the audio. After calculation by the national standard A sound level weighting algorithm, the final A-weighted sound pressure level that meets the sensitivity of the human ear is obtained. The algorithm is as follows:

[0085] Due to the different sensitivities of digital microphones, the time domain sampling point values ​​collected by the microphone are used to calculate the effective sound pressure in the following way:

[0086]

[0087] Among them, P′ e Indicates the effective sound pressure level, the unit is pa, P eIt represents the instantaneous sound pressure collected by the microphone within a certain period of time (in Pascal Pa), x is the size of the signal value converted by the microphone (after standardization, the value ranges from -1 to 1), N is the number of time domain sampling points, that is, the number of effective values ​​per unit time, P ref Represents the reference sound pressure, which is generally 2*10^(-5) Pascal in air, P AOP is the maximum sound pressure level of the hardware microphone (in dB), and then the effective sound pressure level P′ is obtained e Substituting into the equivalent sound pressure level formula, we finally get:

[0088]

[0089] Among them, SPL (sound pressure level) represents the sound pressure level, the unit is db. For example, the maximum sound pressure level PAOP of a certain digital microphone is 120dB, and the RMS value of the currently collected sound data is -22.2dB, then the current sound pressure level is -22.2dB+120dB=97.8dB. Then it is decomposed into octave sound pressure levels through Fourier transform, and corrected by comparing with the A-weighted value, as shown in Table 1 below:

[0090] Table 1 A-weighted octave comparison table

[0091] Center frequency 125 250 500 1000 2000 4000 Octave 90 90 90 90 90 90 A weighted value -16.1 -8.6 -3.2 0 1.2 1

[0092] Original Octave

[0093] After A-weighting correction:

[0094]

[0095] In this embodiment, the multivariate evidence data and the inverted audio sound pressure level data of the data storage unit can be subsequently optimized and integrated, and ultimately complete data for far-field noise event forensics based on drones can be obtained, including but not limited to regional images of noise events, noise occurrence audio, geographic information of noise events, excessive sound pressure levels, regional noise distribution heat maps, etc.

[0096] Example 2

[0097] like Figure 4 As shown, this embodiment provides a method for collecting far-field noise forensics of a drone-mounted sound pickup device, which is provided with the far-field noise forensics system of the drone-mounted sound pickup device in the above-mentioned embodiment 1, and specifically includes the following steps:

[0098] S1: Collect multi-rotor UAV parameter data, build a database of UAV flight parameter data, perform preliminary UAV structure design according to the UAV selection (preliminary setting of background microphone, video acquisition module, far-field noise pickup module, fishing module, vibration reduction module, audio processing module and data storage unit, etc., and detailed parameters will be adjusted later), and replace with ultra-carbon fiber noise reduction blades of matching models.

[0099] In this embodiment, UAV flight parameter experiments and data database construction are carried out at different altitudes (5m-100m, intervals of 5m) and different fishing heights (0m-50m, intervals of 5m). The flight data of the UAV at different altitudes is mainly collected and stored through the background microphone (including but not limited to flight altitude, fishing altitude, wind speed, meteorological conditions, audio files (while unifying the sampling rate, slice duration, data format, number of channels, upper and lower limits of bandwidth and spectrum analysis data results, etc.), and the flight data based on the first background noise signal is accumulated prior knowledge, the sound-collecting structure and material of the noise reduction sound-collecting cover are selected, the flight altitude of the multi-rotor UAV is set, and the fishing altitude of the far-field noise pickup module is set. A parameter database for different flight monitoring scenarios is established, including: the UAV noise is mainly steady noise, according to the actual flight of the UAV In the experiment, the optimal fishing height was set at 20-25m (r2 is the distance from the drone to the pickup microphone, r1 is the distance between the pickup microphone in the noise reduction hood, about 0.3-0.5m) in combination with the sound attenuation algorithm. The flight altitude was set at 50-60m, which can greatly reduce the background noise interference (32-34dB) and improve the effectiveness of audio collection. The wind speed on that day was below 5m / s, there was no rain or snow, the sampling rate was 44800, the duration was 10s (after unified data collection, it was sliced ​​into 10s for analysis and storage), the Wav saving format, the number of channels was 1, and the noise-sensitive bandwidth of the recipient residents was 20hz-20000hz.

[0100] In this embodiment, the sound attenuation algorithm is as follows:

[0101] (1) When the wavelength is much larger than the size of the sound source and in the far field, the sound source can be regarded as a point sound source. Under the far field conditions of the free sound field, the sound power level is used as the evaluation quantity of the point sound source, and the attenuation formula is:

[0102]

[0103] Among them, ΔL represents the attenuation value caused by the increase in distance, the unit is dB, and r represents the distance from the point sound source to the sound receiving point, the unit is m.

[0104] (2) At the distance from the point sound source, the attenuation value from r1 to r2 is expressed as:

[0105] ΔL=20lg(r1 / r2)

[0106] When r2=2r1, ΔL=-6dB, that is, the sound propagation distance of the point sound source increases by 1 times, and the attenuation value is 6dB.

[0107] S11: The sound insulation layer and the sound absorption layer of the noise reduction sound gathering cover are designed by combining the sound insulation and sound absorption materials, and the reflection layer of the noise reduction sound gathering cover is set according to the sound gathering principle;

[0108] In this embodiment, by setting up a drone flight parameter database, collecting the background noise audio of the drone at different altitudes, and using Fourier transform to perform spectrum analysis, such as Figure 5 As shown in the figure, the main background noise sound power spectrum of the drone is obtained, and then the noise reduction material of the noise reduction and sound gathering cover is selected and the sound gathering structure is optimized. The main frequency bands are concentrated in 70-85hz, 150-170hz, and 290-330hz. For steady noise (below 200hz), high-efficiency damping sound insulation felt is selected, and for (above 200hz), melamine sponge plus melamine foam is selected.

[0109] S12: Adjust and optimize the internal reflection focus of the noise reduction sound collecting cover where the pickup microphone is located, amplify the pickup sound, and improve the signal-to-noise ratio;

[0110] In this embodiment, the optimal fishing height is set at 20-25m, and the flight altitude is set at 50-60m. At this time, the sound waves propagating in the far field can be regarded as parallel waves, and the reflection phenomenon on the concave mirror obeys the law of wave reflection. After the sound waves parallel to the main axis are reflected by the concave mirror, the reflected sound waves converge at the focus. Finally, it is determined that the pickup microphone is fixed at the focus of the noise reduction sound collection cover about 0.3-0.5m away from the drone body when there is no flight monitoring task.

[0111] S2: Perform flight monitoring tasks, and collect the second background source noise signal of the drone during the flight monitoring task based on the background microphone;

[0112] In this embodiment, since there are large differences in the environmental interference factors of the drone during each flight monitoring mission (the reference drone parameter database includes but is not limited to flight altitude, fishing altitude, wind speed, meteorological conditions, etc., and the most important sound prints of different noise events collected during the flight monitoring mission), the collected first background noise signal is different from the second background source noise signal.

[0113] In this embodiment, the second background source noise signal data is transmitted to the audio processing module, and the sound power spectrum is obtained by spectrum analysis (short-time Fourier algorithm) to obtain the main frequency band of the second background source noise;

[0114] S3: Perform flight monitoring tasks and collect ground far-field noise signals during flight monitoring tasks based on the pickup microphone;

[0115] In this embodiment, the pickup microphone adopts a supercardioid directional microphone, which can effectively pick up the sound of far-field noise events in the target direction and suppress the interference signal passing through the noise reduction and focusing cover above the microphone.

[0116] S4: Generate an anti-phase signal -ε(n) of the second background source noise signal based on a priori tuned active noise reduction algorithm, and superimpose it with the ground far-field noise signal to generate an anti-noise signal;

[0117] In this embodiment, the recursive least squares adaptive filtering algorithm is improved by using the priori tuning active noise reduction algorithm, aiming to combine the priori knowledge accumulated in the previous experiments to set the weight factor of the filter (hyperparameters include the forgetting factor, step size parameter, filter initial weight and covariance matrix, etc. for initialization strategy adjustment), thereby enhancing the performance of the recursive least squares adaptive filter, and being able to control the input signal more quickly, more accurately and more stably, while retaining the forgetting factor of the algorithm itself will also maintain the predictability of the original algorithm for future signals;

[0118] like Figure 6 As shown, the priori tuning active noise reduction algorithm specifically includes:

[0119] In the case of known filter tap weight coefficients at time n-1, the filter tap weight coefficients at time n are obtained through iterative updating. The audio signal picked up by the pickup microphone is d, the audio signal collected by the background microphone is x, the number of full-time domain sampling points corresponding to the audio file is N, the number of sampling points at a certain moment is n, and the filter tap weight coefficient is M. According to various flight parameters during each flight monitoring, the closest unmanned aerial vehicle parameter database of Example 1 is called for tuning iteration to formulate an initialization strategy. Set hyperparameters such as: the forgetting factor is set to λ as 0-0.5, the step size parameter k is a trade-off between convergence speed and tracking performance, and the value is 0-1, the initial weight of the filter is set to a butterworth filter, the initialization weight is the inverse of the ideal filter, and the value is set to 300, and the covariance matrix is ​​set to the unit matrix of the inverse of the forgetting factor to ensure the accuracy and convergence speed of fixed-point operations, and the parameter is a unit matrix of 2.

[0120] (1) Cost function: Use exponentially weighted sum of squared errors:

[0121] J(n)=∑λ n-i |ε(i)| 2

[0122] Among them, 0<λ<1, which is called the forgetting factor. The purpose of introducing the forgetting factor is to give a larger weight to the errors close to time n and a smaller weight to the errors far from time n, ensuring that the observation data in a certain period of time in the past are "forgotten" so that the filter can work in a stable state.

[0123] (2) Error:

[0124] ε(i)=d(i)-w H (n)*x(i)

[0125] (3) We can take the actual input of the filter as the expected response and bring the error into the complete expression of the weighted sum of squared errors:

[0126] J(n)=∑λ n-i |d(i)-w H (n)*x(i)| 2

[0127] (4) In order to minimize the cost function, the partial derivative of the cost function with respect to the weight w is derived:

[0128] Define R(n), r(n):

[0129]

[0130] (5) Optimal solution:

[0131] w(n)=R -1 (n)r(n)

[0132] (6) Iterative update, using the matrix inversion lemma:

[0133] Define the inverse matrix:

[0134] P(n)=R -1 (n)

[0135] (7) Using the matrix inversion lemma:

[0136]

[0137] (8) k(n) is the gain vector:

[0138]

[0139] (9) Iteration:

[0140] R(n)=λR(n-1)+x(n)x H (n)

[0141] r(n)=λr(n-1)+x(n)d * (n)

[0142] (10) Update weights:

[0143] w(n)=w(n-1)+k(n)ε * (n)

[0144] (11) Update error:

[0145] ε(n)=d(n)-w H (n-1)x(n)

[0146] (12) Using the latest filter tap coefficient M, we can calculate the analog signal w H (n-1)x(n), and finally the analog signal inversion signal is superimposed with the pickup signal d to obtain the anti-noise signal ε(n).

[0147] S6: Convert the anti-noise signal into an audio signal file for storage, and calculate and generate the corresponding simulated sound pressure level.

[0148] The present invention aims to collect evidence for noise nuisance and noise exceeding the standard (such as nighttime construction, decoration during non-decoration periods, square dancing, etc.). This type of noise is relatively loud and has more distinctive sounds. The present invention can allow the drone to fly higher and pick up farther. It uses a passive noise reduction method for high-frequency noise and a priori-tuned active noise reduction algorithm for low-frequency noise, and finally performs multi-source audio and video evidence collection for far-field noise events.

[0149] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for collecting evidence of far-field noise events from a drone-mounted sound pickup device, characterized in that: It is equipped with a far-field noise event forensics system using drone-mounted sound pickup equipment, including: a multi-rotor drone, background microphone, far-field noise pickup module, fishing module, vibration reduction module and audio processing module; The background microphone is arranged on the top of the multi-rotor drone, and is used to collect the background noise of the multi-rotor drone at different flight altitudes and in different scenarios; The multi-rotor drone is provided with a video acquisition module for synchronously acquiring on-site video data, the multi-rotor drone is connected to a fishing module via a vibration reduction module, the fishing module is connected to a far-field noise pickup module, the fishing module is used to adjust the fishing height of the far-field noise pickup module, and the far-field noise pickup module is used to pick up far-field noise at different altitudes; The audio processing module includes a spectrum analysis unit, an active noise reduction unit, a data analysis and conversion unit, and a data storage unit; The spectrum analysis unit is used to perform spectrum analysis on the background noise based on short-time Fourier transform, and transmit the spectrum analysis result to the active noise reduction unit; The active noise reduction unit is used to perform Fourier transform on the background noise and then perform iterative simulation in combination with the prior-tuned active noise reduction algorithm to generate an inverted signal of the background noise signal, achieve phase cancellation by superimposing the time domain and far-field noise signals, obtain effective target noise audio by inverse Fourier transform, and transmit it to the data analysis and conversion unit; The data analysis and conversion unit is used to analyze and convert the effective target noise audio, convert the time domain amplitude into sound pressure level data, and transmit it to the data storage unit; The data storage unit is used to receive and store the analysis and conversion results of the data analysis and conversion unit, and store the video data of the ground far-field noise event collected by the video acquisition module; The method for collecting evidence of far-field noise events from drone-mounted sound pickup equipment includes the following steps: The flight parameter data of the multi-rotor drone is collected at different altitudes and different fishing heights, the first background noise is obtained based on the background microphone, and a parameter database of different flight monitoring scenarios is constructed, and the fishing height of the fishing module is set according to the sound attenuation algorithm, and the flight altitude of the multi-rotor drone is set; The sound attenuation algorithm specifically includes: When the wavelength is much larger than the size of the sound source and in the far field, the sound source is regarded as a point sound source. Under the far field conditions of the free sound field, the sound power level is used as the evaluation quantity of the point sound source, and the attenuation formula is expressed as: Among them, ΔL represents the attenuation value caused by the increase in distance, the unit is dB, r represents the distance from the point sound source to the sound receiving point, the unit is m; At a distance from a point sound source, the attenuation value from r1 to r2 is expressed as: ΔL=20lg(r1 / r2) Where r1 represents the distance between the sound pickup microphone and the noise reduction sound collecting cover, and r2 represents the distance from the drone to the sound pickup microphone. Perform flight monitoring tasks, collect the second background source noise signal of the drone during the flight monitoring task based on the background microphone, transmit the second background source noise signal data to the audio processing module, obtain the sound power spectrum through spectrum analysis, and obtain the frequency band of the second background source noise; Perform flight monitoring tasks and collect ground far-field noise signals during flight monitoring tasks based on microphones; Based on the prior-tuned active noise reduction algorithm, an inverted signal of the second background source noise signal is generated, and superimposed with the ground far-field noise signal to generate an anti-noise signal; The priori tuning of the active noise reduction algorithm specifically includes: Using the filter tap weight coefficient at time n-1, the filter tap weight coefficient at time n is obtained through iterative update. The audio signal picked up by the pickup microphone is d, the audio signal collected by the background microphone is x, the number of full time domain sampling points corresponding to the audio file is N, and the filter tap weight coefficient is M; Each time flight monitoring is performed, the closest parameter in the drone parameter database is called for tuning iteration according to various flight parameters to formulate an initialization strategy; (1) Cost function: Use exponentially weighted sum of squared errors: Among them, 0<λ<1, called the forgetting factor; (2) Error: ε(n)=d(n)-w H (n−1)x(n) Wherein, d(n) represents the ground far-field noise signal sequence picked up by the pickup microphone, and x(n) represents the background noise sequence collected by the background microphone; (3) Taking the actual input of the filter as the expected response, the error is substituted into the complete expression of the weighted error sum of squares: (4) In order to minimize the cost function, the partial derivative of the cost function with respect to the weight w is derived: Define R(n), r(n): (5) Optimal solution: w(n)=R -1 (n)r(n) (6) Iterative update, using the matrix inversion lemma: Define the inverse matrix: P(n)=R -1 (n) (7) Using the matrix inversion lemma: (8) k(n) is the gain vector: (9) Iteration: R(n)=λR(n-1)+x(n)x H (n) r(n)=λr(n-1)+x(n)d * (n) (10) Update weights: w(n)=w(n-1)+k(n)ε * (n) (11) Update error: ε(n)=d(n)-w H (n−1)x(n) (12) Using the latest filter tap coefficient M, we can calculate the analog signal w H (n-1)x(n), and finally the analog signal inversion signal is superimposed with the pickup signal d to obtain the anti-noise signal ε(n); Convert the anti-noise signal into an audio signal file for storage, and calculate and generate the corresponding simulated sound pressure level; The calculation generates the corresponding simulated sound pressure level, specifically including: Among them, P′ e Indicates the effective sound pressure level, P e It represents the instantaneous sound pressure collected within the set time, x is the signal amplitude converted by the microphone, N is the number of time domain sampling points, P ref Indicates the reference sound pressure, P AOP is the maximum sound pressure level of the hardware microphone; The effective sound pressure level P′ e Substituting into the equivalent sound pressure level formula, we get: Here, SPL stands for sound pressure level.

2. The method for collecting evidence of far-field noise events of drone-mounted sound pickup equipment according to claim 1 is characterized in that: The far-field noise pickup module includes a fixing rod, a sound card, a noise reduction and sound focusing cover and a pickup microphone. One end of the fixing rod is connected to the fishing module, and the other end of the fixing rod is connected to the noise reduction and sound focusing cover. The pickup microphone is arranged at the internal reflection focus of the noise reduction and sound focusing cover to pick up far-field noise. The sound card is used to convert the electrical signals collected by the background microphone and the pickup microphone into digital signals. The noise reduction and sound focusing cover is used to reduce the background noise transmitted above the drone and converge the far-field noise transmitted on the ground.

3. The method for collecting evidence of far-field noise events of drone-mounted sound pickup equipment according to claim 2 is characterized in that: The noise reduction and sound gathering cover is provided with a metal shell, a sound insulation layer, a sound absorbing layer and a reflective layer from the outside to the inside. The sound insulation layer is used to isolate the background noise transmitted from the top of the multi-rotor drone, the sound absorbing layer is used to absorb the remaining background noise passing through the sound insulation layer, and the reflective layer is used to reflect far-field sound waves and gather the far-field sound waves at the focus inside the noise reduction and sound gathering cover structure.

4. The method for collecting evidence of far-field noise events of drone-mounted sound pickup equipment according to claim 1 is characterized in that: The fishing module includes a fishing retractable controller, a brushless motor, a battery, a fishing line reel and a fishing line. The fishing retractable controller is electrically connected to the brushless motor, the output shaft of the brushless motor is connected to the fishing line reel, the battery is electrically connected to the brushless motor, one end of the fishing line is connected to the fishing line reel, and the other end of the fishing line is connected to the far-field noise pickup module. The fishing retractable controller controls the output shaft of the brushless motor to rotate forward or reverse through a retractable control signal, drives the fishing line reel to rotate, and adjusts the fishing height of the far-field noise pickup module through the fishing line.

5. The method for collecting evidence of far-field noise events of drone-mounted sound pickup equipment according to claim 1, characterized in that: The multi-rotor drone is provided with ultra-carbon fiber noise reduction blades.

6. The method for collecting evidence of far-field noise events of drone-mounted sound pickup equipment according to claim 1, characterized in that: The background microphone is an omnidirectional microphone, the sound pickup microphone is a unidirectional microphone, and both the background microphone and the sound pickup microphone are provided with windproof cotton.

Images