Group flight bat signal separating and positioning method and system

Abstract

The invention discloses a group flight bat signal separation and positioning method and system, and mainly solves the problem that effective separation and three-dimensional positioning are difficult to realize when group flight bat signals are aliasing in the prior art. According to the implementation scheme, the method comprises the steps that a microphone array is used for receiving group-fly bat signals and preprocessing the group-fly bat signals; estimating the number of bats through the energy distribution in the time-frequency domain; estimating an angle of arrival of the group flight bat signal by adopting an angle measurement algorithm for eliminating interference signals based on iterative projection; based on the estimated angle of arrival of the bat signal, separating the signal by using a frequency domain beamforming algorithm; matching the separated signals by adopting a signal correlation method of circularly rejecting a maximum value, and estimating arrival time difference of each microphone array according to the cross-correlation characteristic of the same bat signal between the arrays; and performing three-dimensional positioning on the group flight bat by combining the arrival angle and the arrival time difference. The method can effectively separate the aliasing signals of the group-flying bats, achieves the three-dimensional positioning of the group-flying bats, and can be used for the design of bionic radars.

Description

Technical Field

[0001] The present invention belongs to the cross - field of radar signal processing and bat acoustics, and particularly relates to a method and system for separating and locating signals of flying bats in groups, which can be used for bat ethology and bionic radar. Background Art

[0002] Bats are animals that rely on echolocation technology to navigate and forage in a completely dark environment. They emit ultrasonic signals and receive the reflected echoes of objects in the environment to accurately perceive the surrounding environment and the position of prey. In the natural scene of bats flying in groups, the acoustic signals simultaneously emitted by multiple individuals are superimposed on each other in space, forming a complex and highly aliased sound field. Bat individuals can fly without interference in this complex interference environment, and their excellent acoustic perception ability is worthy of in - depth study and reference. Studying the group - flying behavior and signal characteristics of bats not only has important value for ecology and animal ethology, but also provides inspiration and theoretical basis for the development of acoustic positioning, multi - target tracking, and radar anti - interference technologies. Due to the aliasing of group - flying bat signals, traditional signal analysis methods face great challenges in signal separation and positioning.

[0003] The current research methods for separating the acoustic signals of flying bats mainly rely on the natural separation characteristics of the signal itself in the time domain. Specifically, researchers usually use fixed microphone arrays or backpack microphone devices to record the echolocation signals of bats. However, whether it is a fixed microphone array or a backpack microphone device, it is difficult to avoid the aliasing of multiple individual signals received at the same time. In 2008, Chiu et al. used 16 fixed microphones arranged in a U-shape on the wall and 3 microphones arranged on the ground in their paper "Flying in silence: Echolocating bats cease vocalizing to avoid sonar jamming" to record the acoustic signals of bats flying in pairs; in 2021, Corcoran et al. used miniature backpack microphones installed on individual bats in their paper "Silence and reduced echolocation during flight areassociated with social behaviors in male hoary bats (Lasiurus cinereus)" and in 2022, Hase et al. used miniature backpack microphones installed on individual bats in their paper "Echo reception in group flight by Japanese horseshoe bats, Rhinolophus ferrumequinum nippon" to directly collect the acoustic signals of flying individuals. The above studies only rely on the natural staggering of bat signals in the time domain to separate them. When the time-frequency aliasing of the sound signals of flying bats is serious, effective separation cannot be achieved, making it difficult to efficiently extract and accurately analyze the signals of flying bats.

[0004] In terms of positioning, current research usually uses a small number of microphones to perform three-dimensional positioning based on the time difference of arrival (TDOA) between microphones. et al. in their paper "No evidence for spectral jamming avoidance in echolocation behavior of foraging pipistrelle bats" used a horizontal T-shaped planar microphone array and the TDOA method for target localization. In 2023, Léna de Framond et al. in their paper "Calibrated microphone array recordings reveal that a gleaning bat emits low-intensity echolocation calls even in open-space habitat" used a planar symmetric star microphone array and also utilized the TDOA method to achieve target localization. These studies can achieve three-dimensional localization of a single bat through the combination of a microphone array and the TDOA method. However, in the scenario of swarming bats, severe signal aliasing will increase the estimation error of TDOA, making it difficult to achieve multi-bat localization under swarming conditions. Summary of the Invention

[0005] Aiming at the deficiencies of the above-mentioned prior art, the present invention proposes a method and system for separating and locating signals of swarming bats, aiming to effectively separate multi-bat aliased signals and achieve three-dimensional localization of bats, providing a new technical means for bat ethology research.

[0006] To achieve the above object, the technical solutions adopted by the present invention include the following:

[0007] 1. The method for separating and locating signals of swarming bats provided by the present invention is characterized by including

[0008] Receiving the signals x m,q (t) of swarming bats by using at least two microphone arrays, and obtaining the l-th frame of aliased signals after amplitude-phase consistency correction, band-pass filtering and frame segmentation processing where l = 1, 2..., L is the frame number, L is the total number of frames, m = 1, 2,..., M is the array number, M is the number of arrays, q = 1, 2,..., Q is the microphone number, and Q is the total number of microphones in each array;

[0009] According to the energy distribution of the l-th frame of aliased signals in the time-frequency domain, estimating the number D of bats in this frame (l) ;

[0010] According to the number D of bats (l) , iteratively projecting to eliminate interference signals, and performing angle measurement on the l-th frame of signals of each array to obtain the angle of arrival AOA of the bats, where

[0011] According to the angle of arrival (AOA) of bats, the frequency-domain beamforming method is used to separate the signals and obtain the signals of each bat in the l-th frame where d = 1, 2,..., D (l) is the bat signal number;

[0012] The signal correlation method of circularly eliminating the maximum value is adopted to gradually match the bat signals of each array and, based on the cross-correlation characteristics of the same bat signal between arrays, estimate the time difference of arrival (TDOA) between arrays of the same bat signal;

[0013] The three-dimensional positions of each bat are calculated by combining the AOA of bats and the TDOA between arrays.

[0014] Furthermore, according to the number of bats D( l ), the interference signals are eliminated by iterative projection, and the signals of the l-th frame of each array are angle-measured to obtain the AOA of bats, including:

[0015] a) Apply the MUSIC algorithm to the signals for angle measurement to obtain the number of spectral peaks P and the corresponding AOAs;

[0016] b) Judge whether the number of spectral peaks P is less than the target number D (l) : If so, execute step c); otherwise, select the AOAs corresponding to the D( l ) spectral peaks with the strongest energy from the P spectral peaks as the AOAs of bats;

[0017] c) Continue to estimate the AOAs of the D -P bat signals masked in the signals (l) by the way of iterative projection to eliminate interference signals, and merge them with the P AOAs obtained in a) to obtain D (l) AOAs of bats.

[0018] Furthermore, the way of continuing to estimate the AOAs of the D -P bat signals masked in the signals (l) by the way of iterative projection to eliminate interference signals includes:

[0019] i) Initialize the known number of spectral peaks to 0;

[0020] ii) Construct an orthogonal projection matrix according to the AOA corresponding to the spectral peak with the strongest energy in the signals and use this matrix to process the signals Perform projection on the covariance matrix to obtain an updated covariance matrix;

[0021] iii) Apply the MUSIC algorithm to the updated covariance matrix for angle measurement, obtain the current number of spectral peaks and the corresponding angles of arrival, and accumulate the newly obtained number of spectral peaks to the known number of spectral peaks;

[0022] iv) Determine whether the known number of spectral peaks is less than the target number D of masking (l) -P: If so, execute v), otherwise, select the D (l) -P angles of arrival corresponding to the spectral peaks with the strongest energy as the angles of arrival of the masked bats;

[0023] v) Extract the angle of arrival corresponding to the spectral peak with the strongest energy in iii) to form an orthogonal projection matrix, update the covariance matrix, and then return to iii).

[0024] Furthermore, the method of constructing an orthogonal projection matrix according to the angle of arrival corresponding to the spectral peak with the strongest energy in the signal and using this matrix to project the covariance matrix of the signal to obtain an updated covariance matrix includes:

[0025] Extract the angle of arrival AOA corresponding to the strongest spectral peak in the signal and construct the steering vector a of this signal; Construct the projection matrix P using the steering vector a:

[0026] where I is the identity matrix, and (·)

[0027]

[0028] represents the conjugate transpose; H denotes the conjugate transpose;

[0029] Project the covariance matrix R of the signal using the projection matrix P to obtain an updated covariance matrix R = PRP n = PRP H .

[0030] 2. Based on the same inventive concept, the present invention also proposes a system for separating and positioning signals of flying bats in groups, which includes:

[0031] A signal receiving and preprocessing module, configured to receive signals of flying bats in groups using at least two microphone arrays, perform amplitude-phase consistency correction, band-pass filtering, and frame segmentation processing on the signals, and generate normalized aliased signal frames;

[0032] A bat number estimation module, configured to perform time-frequency analysis on the normalized aliased signal frames to estimate the number of bats included in the signal frames;

[0033] The Angle of Arrival (AOA) estimation module is used to iteratively project and eliminate interference signals according to the estimated number of bats, and obtain the Angle of Arrival (AOA) of each bat signal.

[0034] The signal separation module is used to separate each bat signal by using the frequency-domain beamforming method according to the Angle of Arrival (AOA) of each bat signal.

[0035] The Time Difference of Arrival (TDOA) estimation module is used to match the bat signals separated by each array, and estimate the Time Difference of Arrival (TDOA) of the matching signals reaching each array based on the cross-correlation characteristics of the same bat signal between arrays.

[0036] The three-dimensional positioning module is used to jointly estimate the three-dimensional spatial position of each bat based on the Angle of Arrival (AOA) and the Time Difference of Arrival (TDOA).

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] First, the present invention uses an angle measurement method of iterative projection to eliminate interference signals. First, the Multiple Signal Classification (MUSIC) algorithm is used to estimate the angle of strong interference signals. An orthogonal projection matrix is constructed based on this angle, and the signal is projected into the subspace orthogonal to the strong signal by using this matrix to eliminate the interference of this strong signal, realizing the angle measurement of weak signals under the masking of strong signals. Further, all strong signals are gradually eliminated through iteration, improving the angle measurement performance of weak targets in an environment with strong and weak intertwined signals.

[0039] Second, the present invention uses the frequency-domain beamforming method to separate the time-frequency aliased bat signals. An optimal beamforming weight vector is constructed based on the Angle of Arrival (AOA), an airspace filter pointing to each bat signal is formed, and the signal is weighted and filtered in the frequency domain, thereby separating the time-frequency aliased broadband bat signals and improving the separation ability of complex aliased signals in a multi-bat environment.

[0040] Third, the present invention uses a signal correlation method of circularly eliminating the maximum value. Based on the normalized correlation coefficients of the bat signals of each array, a correlation coefficient matrix is constructed, the maximum value is found and the corresponding rows and columns are set to zero, reducing the influence of the cross-correlation between strong and weak signals on the auto-correlation of weak signals while realizing the matching of strong signals. Further, through circulation, the correlation matching of all bat signals is gradually realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of the method for separating and positioning signals of flying bats provided in Embodiment 1 of the present invention.

[0042] Figure 2 is a block diagram of the system for separating and positioning signals of flying bats provided in Embodiment 2 of the present invention.

[0043] Figure 3It is the time-frequency and waveform simulation diagram of the signals received by the microphone array in the present invention when two bats are flying.

[0044] Figure 4 It is the simulation diagram of the AOA estimation result of the bat signals with time-frequency aliasing and strong and weak interweaving in the present invention.

[0045] Figure 5 It is the simulation diagram of the result of separating the bat signals with time-frequency aliasing and strong and weak interweaving in the present invention.

[0046] Figure 6 It is the simulation diagram of reconstructing the flight trajectories of two bats and comparing them with the actual flight trajectories in the present invention. Detailed implementation manners

[0047] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0048] Embodiment 1, Method for Separating and Locating Signals of Flying Bats in Groups

[0049] Refer to Figure 1 , and the implementation steps of this example include the following:

[0050] Step 1, Reception and preprocessing of signals of flying bats in groups.

[0051] With the rapid development of radar technology, traditional radar systems face many challenges in multi-target detection, noise suppression, and adaptability to complex environments. In this context, bionics provides a new research direction for the innovation of radar technology, especially the bat echolocation system from nature. Bats are animals that rely on echolocation for navigation and foraging in a completely dark environment. They emit ultrasonic signals and receive the reflected echoes of objects in the environment to accurately perceive the surrounding environment and the position of prey. In the natural scenario of bats flying in groups, the acoustic signals simultaneously emitted by multiple individuals are superimposed on each other in space, forming a complex and highly aliased sound field. Bat individuals can fly without interference in this complex interference environment, and their excellent acoustic perception ability is worthy of in-depth research and reference. This interdisciplinary research not only has important value for ecology and animal behavior, but also provides inspiration and theoretical basis for the development of acoustic positioning, multi-target tracking, and radar anti-jamming technologies. Existing bat data acquisition and analysis methods still cannot completely overcome the key technical bottlenecks in the research of bat echolocation in groups, mainly manifested as the serious aliasing of the echolocation signals of group bats, making it difficult to achieve effective separation and accurate positioning. The method and system for separating and locating signals of flying bats in groups proposed by the present invention aim to effectively separate the overlapping signals of multiple bats and achieve three-dimensional positioning of bats.

[0052] In order to analyze and accurately locate the signals of flying bats in groups, it is first necessary to collect and preprocess the high-frequency ultrasonic signals they emit. The specific implementation includes the following:

[0053] (1.1) Receive swarming bat signals using at least two microphone arrays:

[0054] When using a microphone array for spatial acquisition of ultrasonic waves from swarming bats, in order to obtain the target azimuth and elevation angle information, a planar array or a three-dimensional array is usually adopted. To achieve the positioning of the target, at least two microphone arrays need to be deployed; increasing the number of arrays can fuse more spatial information and further improve the positioning accuracy.

[0055] Since the frequency of bat ultrasonic signals is relatively high, usually above 20 kHz to 120 kHz, in order to ensure that the array can accurately capture the signal characteristics of bats, high-sensitivity ultrasonic microphones need to be selected, whose frequency response range should cover the bat ultrasonic frequency band, and a sampling frequency much higher than the Nyquist frequency should be adopted to ensure signal integrity and reduce sampling errors. Finally, multi-channel received signals are obtained, and their mathematical expression is as follows:

[0056] x m (t) = [x m，1 (t),..., x m，q (t), …, x m，Q (t)] T

[0057] In the formula, m = 1, 2,..., M is the array number, M is the number of microphone arrays, q = 1, 2..., Q is the microphone number, and Q is the number of microphones in each array.

[0058] In this embodiment, the highest main frequency of bats is about 72 kHz, the sampling rate of the array is set to 250 kHz to meet the acquisition requirements of high-frequency signals, and two microphone arrays are used to achieve the positioning of the target.

[0059] (1.2) Perform amplitude and phase consistency correction on the received signals x m,q (t) of each channel of the array:

[0060] In an actual system, there are differences in the amplitude and phase responses of the microphones, front-end amplifier circuits, and analog-to-digital conversion channels of each channel of the array. These differences will seriously affect the performance of subsequent array processing algorithms such as spatial spectrum estimation and beamforming. Therefore, before performing array signal processing, it is necessary to perform consistency correction on the amplitude and phase of the array received signal x m,q (t) to make the amplitude-phase characteristics of all channel received signals consistent. The specific implementation includes the following:

[0061] (1.2.1) Estimate the correction coefficients of each channel

[0062] To correct the amplitude and phase differences of different channels in the microphone array, in a noise-free experimental environment, a standard narrowband signal source with a frequency of f0 is used to irradiate the microphone array. In this embodiment, f0 = 72 kHz is selected as the reference frequency of the calibration signal. After each channel of the microphone array synchronously collects this narrowband signal, the received signal of each channel is obtained. Perform a fast Fourier transform FFT on the received signal of each channel, and extract the complex spectral value Z q (f0). Select the first channel as the reference, and its complex spectral value is denoted as Z1(f0), and calculate the calibration coefficient G q :

[0063]

[0064] (1.2.2) Correct the received signal x m,q (t)

[0065] Perform a fast Fourier transform FFT on the received signal x m,q (t) of each channel of the array to obtain the frequency-domain signal X m,q :

[0066] X m,q (f) = FFT{x m,q (t)}

[0067] Perform amplitude-phase consistency correction on the frequency-domain signal X m,q (f) to obtain the corrected frequency-domain signal X' m,q :

[0068] X' m,q (f) = G q ·X m,q (f)

[0069] where m = 1, 2 is the array number, G q is the calibration coefficient of each channel, q = 1, 2..., Q is the microphone number, and Q is the number of microphones in each array.

[0070] Convert X' m,q (f) back to the time domain through the inverse fast Fourier transform IFFT to obtain the corrected time-domain signal x' m,q :

[0071] x' m,q (t) = IFFT{X' m,q (f)}

[0072] (1.3) Perform band-pass filtering on the corrected signal x' m,q (t)

[0073] To retain the signals within the echolocation frequency band of bats and filter out the interference and noise outside the echolocation signal frequency band, it is necessary to perform band-pass filtering on the corrected signal x′ m,q (t).

[0074] In this embodiment, the greater horseshoe bat is selected, and the echolocation signal frequency range of this bat population is 55 kHz - 72 kHz. According to this frequency range, a finite impulse response (FIR) band-pass filter is designed with the following parameters: passband cut-off frequencies of 50 kHz - 75 kHz, stopband cut-off frequencies of 40 kHz and 80 kHz, sideband attenuation of 60 dB, and passband ripple of 0.1 dB. Use this filter to filter the received signals of each channel to obtain the filtered signal x′ m ′ ,q (t).

[0075] (1.4) Perform frame segmentation on the filtered signal x′ m ′ ,q (t).

[0076] Since the duration of the received signal is relatively long, containing multiple echolocation pulses of each bat, and the bat is always in a moving state, resulting in continuous changes in the pulse emission position, it is necessary to perform frame segmentation on the signal.

[0077] The echolocation signal of the bat is pulse-like, and its signal duration and pulse interval can be used to guide frame segmentation. For the greater horseshoe bat in this embodiment, the duration of its echolocation signal is approximately 10 ms, and the pulse interval is approximately 25 ms. Based on these two parameters, with every 35 ms as a frame, perform frame segmentation on the filtered signal x′ m ′ ,q (t) to generate the frame signal where l = 1, 2..., L is the frame number, and L is the total number of frames.

[0078] Write the l-th frame signal after amplitude-phase consistency correction and band-pass filtering in matrix form

[0079]

[0080] where m = 1, 2 is the array number, q = 1, 2..., Q is the microphone number, and Q is the number of microphones in each array.

[0081] Step 2, estimate the number of bats.

[0082] Before separating and locating the target signals of bats, it is necessary to first determine the number of bat individuals in each frame of the signal. In each frame of the signal to be processed , the number D (l) of bat individuals can be estimated through the time-frequency diagram. The specific implementation includes the following:

[0083] (2.1) Obtain the time-frequency distribution of the signal to be processed through time-frequency analysis methods, such as short-time Fourier transform (STFT), Wigner-Ville distribution transform, and wavelet transform, etc., and generate a time-frequency diagram. The time-frequency distribution is obtained, and a time-frequency diagram is generated.

[0084] In this embodiment, STFT transform is used but not limited to it to calculate the time-frequency distribution.

[0085]

[0086] Among them, w(τ - t) is a window function, such as Gaussian window, Hanning window, and Hamming window, etc. In this embodiment, Hamming window is used.

[0087] (2.2) In the generated time-frequency diagram, count the number of independent energy concentration regions, which is the number of bats D. (l) .

[0088] Due to the differences in frequency, time domain, and energy distribution of different bat signals, they are manifested as multiple independent energy concentration regions in the time-frequency diagram. Therefore, by counting the number of these energy concentration regions, the number of bats D (l) simultaneously emitting sounds can be estimated.

[0089] Step 3, estimate the angle of arrival (AOA) of the bat signal.

[0090] The present invention uses the multiple signal classification (MUSIC) method based on iterative projection to eliminate interference signals to estimate the l-th frame signal of each array for the angle of arrival of each bat signal, including azimuth angle α and elevation angle θ. By iteratively projecting the signal into the signal subspace orthogonal to the strong interference signal for multiple times, the masking of the strong interference signal on the weak signal is eliminated, so that the angle of arrival of the weak bat target can be accurately estimated. The specific implementation includes the following:

[0091] (3.1) The MUSIC method obtains the spatial spectrum:

[0092] (3.1.1) Calculate the covariance matrix R of the l-th frame signal of the m-th array:

[0093]

[0094] Among them, Ε[·] represents the mathematical expectation operation, and (·) H represents the conjugate transpose;

[0095] (3.1.2) Perform eigenvalue decomposition on R to extract the noise subspace eigenvector matrix E n :

[0096]

[0097] Among them, E s and E n represent the eigenvector matrices of the signal subspace and the noise subspace respectively, and Λ s and Λ n are the corresponding eigenvalue matrices;

[0098] (3.1.3) Based on the fact that the eigenvector matrix E n of the noise subspace is orthogonal to the signal steering vector, utilize this orthogonality to construct the spatial spectrum search function P(α,θ):

[0099]

[0100] Among them, is the steering vector of the array, and each steering vector corresponds to an arrival angle. (x q , y q , z q ) are the coordinates of the q-th microphone in the array with the center of the microphone array as the origin, λ is the wavelength of the bat signal;

[0101] (3.1.4) Determine the spatial spectrum according to the spectrum search function P(α,θ), that is, when the steering vector a(α,θ) points to the true arrival direction of the bat signal, the spatial spectrum function corresponding to this angle will show an obvious peak. By traversing all possible angles within the entire search space, obtain the complete spatial spectrum;

[0102] (3.2) Count the number P of peaks in the spatial spectrum, and select an appropriate angle measurement strategy according to the relationship between the number P of peaks and the number D (l) of bats:

[0103] When the number of peaks P is greater than or equal to the number of bats D (l) , select the arrival angles corresponding to the D (l) peaks with the strongest energy from the P peaks as the angle of arrival (AOA) of the bats;

[0104] When the number of peaks P is less than the number of bats D( l ), that is, the masking phenomenon occurs. Continue to use the subsequent steps to estimate the arrival angles of the D (l) -P bat signals that are masked;

[0105] (3.3) Iteratively project to eliminate interference signals and estimate the AOAs of the D (l) -P bat signals:

[0106] (3.3.1) Initialize the known number of peaks to 0;

[0107] (3.3.2) The arrival angle (α max , θ max ) corresponding to the spectral peak with the strongest energy in the spatial spectrum obtained according to (3.1.4) is used to construct the steering vector a(α max , θ max ):

[0108]

[0109] where (x q , y q , z q ) are the coordinates of the q-th microphone in the array with the center of the microphone array as the origin, and λ is the wavelength of the bat signal;

[0110] (3.3.3) According to the steering vector a(α max , θ max ), the orthogonal projection matrix P orthogonal to it is obtained:

[0111]

[0112] where I is the identity matrix of P×P;

[0113] (3.3.4) The signal covariance matrix R is projected using the orthogonal projection matrix P to obtain the updated covariance matrix R n :

[0114] R n = PRP H ;

[0115] (3.3.5) Apply the MUSIC algorithm to the updated covariance matrix for angle measurement, obtain the current number of spectral peaks and the corresponding arrival angles, and accumulate the newly obtained number of spectral peaks to the known number of spectral peaks;

[0116] (3.3.6) Determine whether the known number of spectral peaks is less than the masking target number D (l) - P:

[0117] If so, execute (3.3.7), otherwise, select the arrival angles corresponding to the D (l) - P spectral peaks with the strongest energy as the arrival angles of the masked bats;

[0118] (3.3.7) Extract the arrival angles corresponding to the spectral peaks with the strongest energy in (3.3.5) to construct an orthogonal projection matrix, update the covariance matrix, and then return to (3.3.5).

[0119] Step 4: Use the frequency-domain beamforming algorithm to separate each bat signal from the aliased signal.

[0120] After knowing the angle of each bat signal, an airspace filter can be constructed to point to the direction of each bat signal and suppress the signals in other directions, so as to separate each signal. In this embodiment, the frequency-domain linearly constrained minimum variance (LCMV) beamforming algorithm is selected to separate the aliased signals. The specific implementation is as follows:

[0121] (4.1) Calculate the frequency-domain covariance matrix:

[0122] (4.1.1) Transform the l-th frame signal of the m-th array to the frequency domain through the fast Fourier transform (FFT) to obtain the frequency-domain signal

[0123]

[0124] Express the frequency-domain signal using the data at all frequency points as:

[0125]

[0126] where f j is the frequency at the j-th frequency point, j = 1, 2,..., J is the frequency-point number, and J is the total number of frequency points;

[0127] (4.1.2) Calculate the covariance matrix of the frequency-domain signal

[0128]

[0129] (4.2) Calculate the optimal beamforming weight vector:

[0130] (4.2.1) For the d-th bat signal of the m-th array, calculate the frequency-domain steering vector a(f j , α j , θ m,d , θ m,d ):

[0131]

[0132] where α m,d is the azimuth angle of the d-th bat signal estimated by the m-th array, θ m,d is the elevation angle of the d-th bat signal estimated by the m-th array, m = 1, 2 is the array number, d = 1, 2,..., D (l) is the bat-signal number, D (l) is the number of bat signals, (x q , y q , z q) is the coordinate of the q-th microphone in the array with the center of the array as the origin. q = 1, 2,..., Q is the microphone number, and Q is the number of microphones in each array. λ j is the wavelength corresponding to the frequency f j ;

[0133] (4.2.2) Calculate the frequency f j of the interference signal and the corresponding frequency-domain steering vector a(f j , α m,i , θ m,i ):

[0134]

[0135] where α m,i is the azimuth angle of the i-th interference signal estimated by the m-th array, and θ m,i is the elevation angle of the i-th interference signal estimated by the m-th array. i = 1, 2,..., I is the interference signal number, and I is the total number of interference signals;

[0136] (4.2.3) Form the interference signal steering vector matrix A j using the frequency-domain steering vector a(f m,i , α m,i , θ I ) in the direction of the desired interference signal suppression:

[0137] A I = [a(f j , α m,1 , θ m,1 ),..., a(f j , α m,i , θ m,i ),..., a(f j , α m,I , θ m,I )];

[0138] (4.2.4) According to the frequency-domain steering vector a(f j ) at the frequency f j of the d-th bat signal of the m-th array and the interference signal steering vector matrix A m,d m,d I , form the frequency-domain steering vector matrix

[0139]

[0140] (4.2.5) For the d-th bat signal of the m-th array, calculate the frequency f according to the frequency-domain covariance matrix and the frequency-domain steering vector matrix ​j Optimal weights for corresponding frequency-domain LCMV beamforming

[0141]

[0142] where (·) -1 is the inverse operation;

[0143] (4.3) Separate each bat signal:

[0144] (4.3.1) Use the weight vector to weight each frequency point of the received signal to obtain the separated signal at each frequency point

[0145]

[0146] (4.3.2) Obtain the frequency-domain representation of the separated signal based on the separated signal at each frequency point

[0147]

[0148] (4.3.3) Perform the inverse fast Fourier transform on the frequency-domain representation of the d-th bat signal of the m-th array to be separated, to obtain its time-domain representation where m = 1, 2 is the array number, and d = 1, 2,..., D

[0149]

[0150] is the bat signal number; (l) for the bat signal number;

[0151] (4.4) Based on the arrival angles of all bat signals, calculate the optimal weight vector for each bat signal and perform this beamforming to separate all bat signals in each frame of signal of each array.

[0152] Step 5, estimate the time difference of arrival TDOA of each bat signal arriving at each array.

[0153] To achieve 3D positioning of bat signals, in addition to the angle of arrival AOA, it is also necessary to estimate the time difference of arrival TDOA of the signal arriving at each array. Since the signals received by the arrays come from multiple bat individuals, it is first necessary to identify and match the signals from the same bat in different arrays, and then use the cross-correlation characteristics between the matched signals to calculate their time difference of arrival TDOA. The specific implementation includes the following:

[0154] (5.1) Match the signals from the same bat:

[0155] In a multi-target scenario, the received signals often come from multiple different bats. It is necessary to associate the signals separated from each array to match the signals from the same bat. Based on the correlation of signals from the same bat between arrays, a signal association method of circularly removing the maximum value is adopted to gradually match the bat signals of each array. The specific implementation is as follows:

[0156] (5.1.1) Calculate the normalized cross-correlation coefficient c between the bat signals of each array r,n :

[0157]

[0158] where is the r-th signal separated from array 1, r = 1, 2,..., D (l) is the number of the signal separated from array 1, is the n-th signal separated from array 2, n = 1, 2,..., D (l) is the number of the signal separated from array 2, D (l) is the number of bats in this frame of signals, respectively represent the time-domain means of the signals separated from the first array and the second array.

[0159] (5.1.2) According to the normalized cross-correlation coefficient c between the bat signals of each array r,n , form a normalized cross-correlation coefficient matrix C of D (l) ×D (l) :

[0160]

[0161] (5.1.3) Find the maximum value in the normalized cross-correlation coefficient matrix C. The subscripts of the maximum value are the signal numbers separated from the two arrays that belong to the same bat. Denote the signals corresponding to these two numbers as the matching signals y 1,k (t) and y 2,k (t), k = 1, 2,..., D (l) is the bat number, D (l) is the number of bats in this frame of signals;

[0162] (5.1.4) Adopt the method of gradually removing the matched signals to reduce the interference of bats with stronger calls on the matching process. Specifically, each time the maximum value in the current matrix is found, record its subscript as a matching result, and set the row and column where the subscript is located to zero, then continue to find the new maximum value in the matrix, and repeat this process until all signal associations and matches are completed;

[0163] (5.2) The cross-correlation method is used to estimate the arrival time difference TDOA of bat signals arriving at different arrays:

[0164] (5.2.1) According to the matching signal y of the kth bat 1,k (t) and y 2,k (t) Calculate the cross-correlation function R k (τ):

[0165]

[0166] The cross-correlation function R k The τ value corresponding to the peak value of (τ) is the arrival time difference TDOA of the kth bat signal arriving at the two arrays, denoted as t k ,k=1,2,...,D (l) Number the bats, D (l) is the number of bats in the frame signal;

[0167] (5.2.2) Perform the process (5.2.1) for all matching signals to obtain the arrival time difference TDOA of each bat’s signal arriving at each array.

[0168] Step 6: Estimate the bat position based on the angle of arrival (AOA) and the time difference of arrival (TDOA).

[0169] The principle of TDOA and AOA joint positioning is to determine the target position by combining TDOA and AOA measurement equations. From a geometric point of view, the target position is determined by the intersection of the time difference positioning hyperboloid and the angle positioning ray. Solving this intersection can achieve three-dimensional positioning of the bat. The specific implementation includes the following:

[0170] (6.1) The azimuth angle α of the kth bat signal estimated by the mth array m,k and the pitch angle θ m,k , construct the angle information matrix γ m,k :

[0171]

[0172] Among them, γ m,k is the angle information matrix, m=1,2 is the array number, k=1,2,...,D (l) Number the bats, γ 1,k and γ 2,k denote the angle information matrix of the first array and the angle information matrix of the second array respectively, (·) T represents transpose;

[0173] (6.2) According to the angle information matrix γ of the first array 1,k , the angle information matrix of the second array γ 2,kand the central position coordinates s of each microphone array m =[x m ,y m ,z m T , construct the first angle coefficient matrix and the second angle coefficient matrix

[0174]

[0175] (6.3) According to the time difference of arrival t of the k-th bat signal at each array k , calculate the difference in arrival distance r of it at each array k :

[0176] r k =t k *c

[0177] where c is the propagation speed of the bat signal;

[0178] (6.4) According to the azimuth angle α 1,k and elevation angle θ 1,k of the k-th bat signal estimated by array 1, construct the unit vector b:

[0179] b=[cosθ 1,k cosα 1,k ,cosθ 1,k sinα 1,k ,sinθ 1,k T ;

[0180] (6.5) According to the difference in arrival distance r of the k-th bat signal at each array k , the central position coordinates s of each microphone array m =[x m ,y m ,z m T and the unit vector b, construct the first distance coefficient matrix h r,k and the second distance coefficient matrix G r,k :

[0181]

[0182] G r,k =[2(s1 - s2 - r k b) T

[0183] where ||·|| represents the norm, s1 = [x1, y1, z1] T ​​​​is the central position coordinate of the first array, s2 = [x2, y2, z2] T is the central position coordinate of the second array;

[0184] (6.6) Combine the first angle coefficient matrix with the second angle coefficient matrix the first distance coefficient matrix h r,k and the second distance coefficient matrix G r,k to construct the first coefficient matrix h k and the second coefficient matrix G k :

[0185]

[0186]

[0187] (6.7) According to the first coefficient matrix h k and the second coefficient matrix G k , calculate the position coordinate u of the k-th bat k = [x k , y k , z k T :

[0188]

[0189] (6.8) For the angle of arrival AOA and time difference of arrival TDOA of all bat signals, repeat the above steps (6.1) to (6.7) to achieve the position estimation of each bat, and reconstruct the motion trajectory of each bat according to the positions of each bat at different times.

[0190] Embodiment 2, Swarming Bat Signal Separation and Localization System

[0191] Based on the same inventive concept, the present invention also proposes a swarming bat signal separation and localization system, as Figure 2 shown, which includes:

[0192] A signal reception and preprocessing module for receiving swarming bat signals using at least two microphone arrays, and performing amplitude-phase consistency correction, band-pass filtering and frame segmentation processing on them to generate normalized aliased signal frames;

[0193] A bat number estimation module for performing time-frequency analysis on the normalized aliased signal frames to estimate the number of bats contained in the aliased signal frames;

[0194] An angle of arrival AOA estimation module for iteratively projecting and eliminating interference signals according to the estimated number of bats to obtain the angle of arrival AOA of each bat signal;

[0195] ​A signal separation module, which is used to separate the signals of each bat by using the frequency-domain beamforming method according to the angle of arrival (AOA) of each bat signal;

[0196] A time difference of arrival (TDOA) estimation module, which is used to match the bat signals separated by each array, and estimate the time difference of arrival (TDOA) of the matched signals at each array based on the cross-correlation characteristics of the same bat signal between arrays;

[0197] A three-dimensional positioning module, which is used to estimate the three-dimensional spatial position of each bat according to the angle of arrival (AOA) and the time difference of arrival (TDOA).

[0198] The technical effects of the present invention will be further described below in combination with simulation experiments.

[0199] 1. Simulation conditions:

[0200] Two 8×8 uniform planar microphone arrays are used to receive the echolocation signals from bats. The echolocation signals emitted by bats contain a constant frequency band and a downward frequency modulation band. The speed of sound c = 340 m / s, the center frequency is about 68000 Hz, and the element spacing is half a wavelength.

[0201] Hardware environment: The CPU is Intel(R) Core(TM) i9-14900HX, the main frequency is 2.20 GHz, the memory is 32.0 GB, and the 64-bit operating system.

[0202] Software environment: Microsoft windows 11 Home Chinese version, MATLAB 2024a simulation software.

[0203] Simulation scenario: The position of array 1 is (0,0,0), and the position of array 2 is (0,0.5,0), with the unit of m. Two bats 1 and 2 fly towards the microphone array, and the echolocation calls emitted by each bat are received by the microphone array, where:

[0204] The constant frequency of the call of bat 1 is 68196 Hz, the bandwidth of the frequency modulation band is 10 kHz, and the amplitude is 0.2;

[0205] The constant frequency of the call of bat 2 is 68141 Hz, the bandwidth of the frequency modulation band is 10 kHz, and the amplitude is 9.6.

[0206] The time taken for bat 1 to fly uniformly and turn from the position (3, -0.58, -0.91) to the position (1.61, -1.11, -0.51) is 1.16 s, and it emits 29 sounds during the flight;

[0207] The time taken for bat 2 to fly uniformly and turn from the position (3, -0.7, -0.21) to (0.44, -0.86, -0.67) is 1.17 s, and it emits 41 sounds.

[0208] The constant frequency band of the two bat signals is 10 ms each, and the frequency modulation band is 2 ms.

[0209] II. Simulation Content

[0210] Simulation 1: When simulating the simultaneous flight of two bats, the time-frequency diagram and waveform diagram of the signals received by the microphone channels at the origin of microphone array 1 are shown as Figure 3 follows. Figure 3 (a) is the time-frequency diagram of the received signal, Figure 3 (b) is the corresponding time-domain waveform diagram. It can be seen from Figure 3 this that when two bats fly simultaneously, there is an obvious time-frequency aliasing phenomenon in the signals received by the microphone array.

[0211] Simulation 2: For a frame of signals with time-frequency aliasing and obvious energy difference between two signals in Simulation 1, the angle is measured using the AOA algorithm for eliminating strong interference signals based on iterative projection in the present invention. The results are shown as Figure 4 follows. Among them:

[0212] Figure 4 (a) is the time-frequency diagram of this frame of signals, including weak signal 1 and strong signal 2. It can be visually estimated from this that the number of signals is 2;

[0213] Figure 4 (b) is the spatial spectrum of this frame of signals obtained using the MUSIC algorithm. Only a single spectral peak can be observed in this spatial spectrum diagram, indicating that the strong signal masks the weak signal, resulting in the weak signal being unable to be effectively identified.

[0214] Figure 4 (c) is the spatial spectrum estimated using the AOA estimation algorithm for eliminating strong interference signals based on iterative projection in the present invention. The spatial spectrum successfully estimates the angle of arrival (AOA) of the weak signal.

[0215] From Figure 4 the results, it shows that the present invention can effectively estimate the angle of the weak signal under the masking of the strong signal.

[0216] Simulation 3: For a frame of signals with time-frequency aliasing and obvious energy difference between two signals in Simulation 2, the effective separation of the signals is achieved using the frequency-domain LCMV beamforming algorithm adopted in the present invention, and the time-frequency diagram after separation is plotted. The results are shown as Figure 5 follows. Among them, Figure 5 (a) is the separation result of signal 1, Figure 5 (b) is the separation result of signal 2. It can be seen from Figure 5 this that the present invention can effectively separate each bat signal.

[0217] Simulation 4: Based on two microphone arrays, using the algorithm for estimating the three-dimensional position of bats according to the Angle of Arrival (AOA) and Time Difference of Arrival (TDOA) adopted in the present invention, estimate the three-dimensional position of flying bats and reconstruct the flight trajectories of two bats. The results are as Figure 6 shown. This figure compares the actual flight trajectories of two bats with the trajectories reconstructed by the algorithm. It can be seen from Figure 6 this that the present invention can accurately reconstruct the motion trajectories of flying bats.

[0218] In summary, a method for separating and locating signals of flocking bats proposed by the present invention can effectively estimate the AOA of bat signals with a large difference in signal strength between strong and weak signals and successfully separate the signals of different individual bats; based on the cooperation of two microphone arrays, the motion trajectories of flying bats can be accurately reconstructed, providing important technical support for bat ethology and bionic radar research.

[0219] The above description is only two specific examples of the present invention and does not constitute any limitation to the present invention. Obviously, for professionals in the field, after understanding the content and principle of the present invention, various modifications and changes in form and details may be made without departing from the principle and structure of the present invention. However, these corrections and changes based on the idea of the present invention are still within the scope of protection of the claims of the present invention.

[0220] It should be noted that the step numbers in the specification and claims of the present invention are only for clear description of the implementation embodiments of the present invention for easy understanding, and their sequence numbers are not limited.

Claims

1. A method for separating and locating signals of swarming bats, characterized in that Including: Receiving the signals x of swarming bats using at least two microphone arrays m,q (t), and obtaining the l-th aliased signal after amplitude-phase consistency correction, band-pass filtering and framing processing where l = 1, 2..., L is the frame number, L is the total number of frames, m = 1, 2,..., M is the array number, M is the number of arrays, q = 1, 2,..., Q is the microphone number, and Q is the total number of microphones in each array; According to the aliasing signal of the l-th frame Estimate the number of bats D in this frame based on the energy distribution in the time-frequency domain (l) ; According to the number of bats D (l) , iteratively project to eliminate interference signals, and perform angle measurement on the l-th frame signal of each array , to obtain the Angle of Arrival (AOA) of the bats, where According to the angle of arrival (AOA) of bats, the frequency-domain beamforming method is used to separate the signals and obtain the signals of each bat in the l-th frame where d = 1, 2,..., D (l) is the bat signal number; Adopt a signal correlation method of circularly removing the maximum value, and gradually match the bat signals of each array Perform matching, and based on the cross-correlation characteristics of the same bat signal between arrays, estimate the time difference of arrival TDOA between arrays of the same bat signal; Calculate the three-dimensional positions of each bat by combining the Angle of Arrival (AOA) of the bat and the Time Difference of Arrival (TDOA) between arrays.

2. The method for separating and positioning the signals of a group of flying bats according to claim 1, characterized in that, The amplitude-phase consistency correction includes: 2a) Irradiate each microphone array with a standard narrowband signal source having a known frequency f0, and synchronously collect the signals z q (t); 2b) Extract the acquired signal z q Complex spectrum value Z at the center frequency of (t) q (f0), select the complex spectrum value of the first channel as Z1(f0) as a reference, and calculate the amplitude-phase correction coefficient 2c) Multiply the correction coefficient G q with the frequency-domain signal of the array received data x m,q (t) to obtain a frequency-domain correction signal and convert it to the time domain to obtain the signal after amplitude-phase consistency correction.

3. A method for separating and locating swarming bat signals according to claim 1, characterized in that, The aliasing signal according to the l-th frame Estimate the number of bats D in this frame based on the energy distribution in the time-frequency domain (l) , including: 3a) Perform time-frequency analysis on the aliased signal of the l-th frame to generate a time-frequency diagram of the signal of this frame; 3b) Identify and count the regions with concentrated energy of independent bat signals in the time-frequency diagram to obtain the number of bats D (l) .

4. A method for separating and locating signals of swarming bats according to claim 1, characterized in that, According to the number of bats D (l) , iteratively project to eliminate interference signals, and perform angle measurement on the l-th frame signal of each array to obtain the angle of arrival (AOA) of the bats, including: 4a) Apply the MUSIC algorithm to the signal to perform angle measurement, obtaining the number of spectral peaks P and the corresponding angle of arrival AOA; 4b) Determine whether the number of spectral peaks P is less than the target number D (l) : If so, execute step 4c). Otherwise, select the arrival angles corresponding to the D( l ) spectral peaks with the strongest energy from the P spectral peaks as the bat arrival angle AOA; 4c) Continuing to estimate the signal by the method of eliminating interference signals through iterative projection The masked D in (l) -P AOA of bat signals, and merging them with the P AOAs obtained in 4a) to obtain (l) The AOA of D bats arriving.

5. A method for separating and locating signals of a group of flying bats according to claim 4, characterized in that, Continue to estimate the signal by the method of eliminating interference signals through iterative projection the masked D in (l) -P AOA of bat signals, including: 5a) Initialize the known number of spectral peaks to 0. 5b) Construct an orthogonal projection matrix according to the arrival angle corresponding to the spectral peak with the strongest energy in the signal and use this matrix to project the covariance matrix of the signal to obtain an updated covariance matrix; 5c) Apply the MUSIC algorithm to the updated covariance matrix for angle measurement, obtain the current number of spectral peaks and the corresponding angles of arrival, and accumulate the newly obtained number of spectral peaks to the known number of spectral peaks. 5d) Determine whether the number of known spectral peaks is less than the target number D to be masked (l) -P: If so, execute 5e). Otherwise, select the D with the strongest energy (l) - The arrival angles corresponding to the P spectral peaks are used as the arrival angles of the masked bats; 5e) Extract the angle of arrival corresponding to the spectral peak with the strongest energy in 5c) to construct an orthogonal projection matrix, update the covariance matrix, and then return to 5c).

6. The method for separating and positioning the signals of a group of flying bats according to claim 5, wherein Construct an orthogonal projection matrix according to the arrival angle corresponding to the spectral peak with the strongest energy in the signal and use this matrix to project the covariance matrix of the signal to obtain an updated covariance matrix, including: 6a) Extract the signal Extract the angle of arrival (AOA) corresponding to the strongest spectral peak in , including the azimuth angle α and the elevation angle θ, and construct the steering vector a(α, θ) of this signal: Among them, (x q , y q , z q ) are the coordinates of the q-th microphone in the array with the center of the array as the origin, and λ is the wavelength of the bat signal; 6b) Use the steering vector a to construct the projection matrix P: where I is the identity matrix, (·) H denotes conjugate transpose; 6c) Project the covariance matrix R of the signal using the projection matrix P to obtain the updated covariance matrix R n = PRP H .

7. A method for separating and locating signals of a group of flying bats according to claim 1, characterized in that, According to the angle of arrival (AOA) of bats, the frequency domain beamforming method is used to separate the signals and obtain the signals of each bat in the l-th frame including: 7a) Perform a fast Fourier transform (FFT) on the signal to obtain a frequency-domain signal and represent it with data at all frequency points: where f j is the frequency at the j-th frequency point, j = 1, 2, ..., J is the frequency point number, and J is the total number of frequency points; 7b) Calculate the covariance matrix of the frequency-domain signal ​ where E[·] represents the mathematical expectation, and (·) H represents the conjugate transpose; 7c) Calculate the frequency f for the d-th bat signal of the m-th array j The corresponding frequency-domain steering vector a(f j , α m,d , θ m,d ): where α m,d is the azimuth angle of the d-th bat signal estimated by the m-th array, θ m,d is the elevation angle of the d-th bat signal estimated by the m-th array, m = 1, 2, …, M is the array number, M is the number of microphone arrays, d = 1, 2, …, D ( l ) is the bat signal number, D ( l ) is the number of bat signals, (x q , y q , z q ) are the coordinates of the q-th microphone in the array with the array center as the origin, q = 1, 2, ..., Q is the microphone number, Q is the number of microphones in each array, λ j is the wavelength corresponding to the frequency f j ; 7d) Calculate the frequency f of the interference signal j The corresponding frequency-domain steering vector a(f j , α m,i , θ m,i ): where α m,i is the azimuth angle of the i-th interference signal estimated by the m-th array, θ m,i is the elevation angle of the i-th interference signal estimated by the m-th array, i = 1, 2, ..., I is the interference signal number, and I is the total number of interference signals. The frequency-domain steering vector a(f j , α m,i , θ m,i ) that is expected to suppress the interference signal direction is used to form the interference signal steering vector matrix A I : A I = [a(f j , α m,1 , θ m,1 ),..., a(f j , α m,i , θ m,i ),..., a(f j , α m,I , θ m,I )] 7e) According to the frequency f of the d-th bat signal of the m-th array j at the frequency domain steering vector a(f j ,α m,d ,θ m,d ) and the interference signal steering vector matrix A I , a frequency domain steering vector matrix is formed 7f) Calculate the frequency f for the d-th bat signal of the m-th array j The optimal weight of the corresponding frequency domain LCMV beamforming where (·) -1 is the inverse operation; 7g) Using the weight vector to weight each frequency point of the received signal separately, and obtain the separated signal at each frequency point Further combine the separated signals at each frequency point to obtain the frequency domain representation of the separated signal 7h) Perform an inverse fast Fourier transform (IFFT) on the frequency-domain components of each bat signal to obtain the time-domain signals of each bat 8. A method for separating and locating signals of a group of flying bats according to claim 1, characterized in that The signal correlation method using cyclic elimination of the maximum value gradually performs matching on the bat signals of each array and estimates the time difference of arrival (TDOA) between arrays of the same bat signal based on the cross-correlation characteristics of the same bat signal between arrays, including: 8a) Calculate the normalized cross-correlation coefficients between each bat signal among the arrays, and construct the correlation coefficient matrix C of D × D (l) ; (l) ×D (l) ; 8b) Find the maximum value in C, regard the bat signals corresponding to the row and column where the maximum value is located as the matched bat signals, and set the corresponding row and column to zero. Loop this step until all bat signals are matched. 8c) Calculate the cross-correlation function for the matched bat signals and estimate the Time Difference of Arrival (TDOA) between arrays based on the position of the cross-correlation peak.

9. A method for separating and locating signals of a group of flying bats according to claim 1, characterized in that, The method for estimating the three-dimensional positions of each bat by combining the Angle of Arrival (AOA) of the bat and the Time Difference of Arrival (TDOA) between arrays includes: 9a) Construct the angle information matrix γ according to the angle of arrival (AOA) of bats in each array m,k : where α m,k and θ m,k represent the azimuth angle and elevation angle of the k-th bat signal estimated by the m-th array respectively, and (·) T denotes transpose. 9b) According to the angle information matrix γ of the first array 1,k , the angle information matrix γ of the second array 2,k and the central position coordinates s of each microphone array m = [x m , y m , z m T , construct the first angle coefficient matrix and the second angle coefficient matrix ​ 9c) Calculate the difference in arrival distances r of the k-th bat signal at each array based on the time difference t of arrival of the k-th bat signal at each array. k k : r k = t k * c where c is the propagation speed of the bat signal; 9d) Azimuth angle α of the k-th bat signal estimated according to array 1 1,k and elevation angle θ 1,k , construct a unit vector b: b = [cosθ 1,k cosα 1,k , cosθ 1,k sinα 1,k , sinθ 1,k T ;​ 9e) According to the arrival distance difference r of the k-th bat signal at each array k , the central position coordinates s of each microphone array m = [x m , y m , z m T and the unit vector b, construct the first distance coefficient matrix h r,k and the second distance coefficient matrix G r,k :​ G r,k = [2(s1 - s2 - r k b) T ​ where ||·|| represents the norm. 9f) Combine the first angular coefficient matrix with the second angular coefficient matrix and the first distance coefficient matrix h r,k and the second distance coefficient matrix G r,k to construct the first coefficient matrix h k and the second coefficient matrix G k : 9g) According to the first coefficient matrix h k and the second coefficient matrix G k , calculate the position coordinates u k = [x k , y k , z k T :​ 10. A system for separating and locating signals of a group of flying bats, characterized in that, Including: A signal reception and preprocessing module, which is used to receive the signals of flying bats using at least two microphone arrays, perform amplitude-phase consistency correction, band-pass filtering and frame segmentation on them, and generate normalized aliased signal frames. A bat number estimation module, which is used to perform time-frequency analysis on the normalized aliased signal frames to estimate the number of bats contained in the signal frames. An Angle of Arrival (AOA) estimation module, which is used to iteratively project and eliminate interference signals according to the estimated number of bats to obtain the Angle of Arrival (AOA) of each bat signal. A signal separation module, which is used to separate each bat signal using the frequency-domain beamforming method according to the Angle of Arrival (AOA) of each bat signal. A Time Difference of Arrival (TDOA) estimation module, which is used to match the bat signals separated by each array and estimate the Time Difference of Arrival (TDOA) of the matched signals arriving at each array based on the cross-correlation characteristics of the same bat signal between arrays. A three-dimensional positioning module, which is used to estimate the three-dimensional spatial positions of each bat by combining the Angle of Arrival (AOA) and the Time Difference of Arrival (TDOA).

Images